Day 2: Gibbons Lecture - Agency Theory in Organizations

The economics literature has long attributed the problems inside of firms to information asymmetry, more specifically ‘hidden actions.’ The principal can not observe the actions of the agent, but can only observe the output, which has noise. The agent proceeds to provide sub-optimal effort levels. This is the classic moral hazard problem. There are three caveats here. One, output production has noise, so our principal can not deduce with certainty the agent’s effort level from output. Second, our principal can contract perfectly on output, so our principal can do everything in his power to create enough incentives through contracts to preempt a shirking worker. Thirdly, our agent has to be risk averse. As a result, our principal’s problem is to write a contract that creates enough incentives for the agent to work hard, while still providing enough insurance for the agent.

However, with the concept of non-contractibles creeping into economics, we are provided with a new option in modeling agency problems. While moral hazard is a serious and authentic problem inside firms, we can not help but wonder how much managers or firm owners think about risks when drawing contracts. Instead, an agency problem can arise simply because the principal is paying for the wrong things. For example, let us assume that a firm is concerned with its output quantity and quality. However, quality is not contractible. As a result, the firm contracts solely on quantity, then the worker is provided with full incentives to boost quantity, all the while ignoring the products’ quality. This is the so called ‘getting what you pay for problem.’ In short, the principal can not align the interests of the agents with their own, simply because the things the principal care about can not be included in the contract. We shall call contracts of these types ‘formal contracts.’ Their key feature is that the firms choose to contract on objective and measurable aspects which are verifiable in court, but often a misalignment of interests arise. Such multi-task models where only a portion of those tasks are contractible were developed by Holmstrom and Milgrom (1991) and Baker (2002).

There is a solution to non-contractibles though. Continuing our example of quantity and quality, a firm can promise in advance to award the worker a bonus if he also produces high quality output. Since quality is not verifiable outside of the relationship, the principal has an incentive to back out of his initial promise. However, a contract such as this could be sustained in the long-run, where the worker chooses to produce high quality output and the principal pays a bonus in every period. In other words, such ‘relational contracts’ are strongly dependent on the likelihood of the principal and agent’s continuous cooperation. If the likelihood of continuous cooperation is high, then a relational contract is much more likely to be sustained. The concept of relational contract was formally introduced by Bull (1987) and then further advanced by Levin (2003).

As you can see, I did not summarize the models that Gibbons introduced. I have written them down somewhere though.