The principal typically retains some decision making authority and partially delegates decisions to the agent
Recent work has examined the costs and the benefits of delegation versus centralization
The literature contains three approaches:
How does centralization affect the agents gathering of information? (AT)
How does centralization affect the agent’s transfer of information? (GK, MM, HR, D)
How does centralization affect the agent’s implementation of tasks? (Z)
This paper will focus on one primary example of a shared decision rule, namely the veto rule. We will compare this to complete delegation.
The manager will have private information about project quality and a proclivity to spend more than the principal, due to private benefits of spending.
Delegation has the advantage that the manager never has an inventive to distort information. However, it has the disadvantage of allowing the manager to take on undesirable projects.
Veto power in the hands of the principal will induce the agent to distort his private information, but will allow the principal to shut down undesirable projects.
Does one of these decision rules dominate the other?
Dessein (2002) says yes. In fact, for reasonable preference differences between the principal and the agent, delegation is superior. More precisely he says that delegation dominates veto unless the divergence in preferences is so large that informative communication is not possible.
This conclusion is unsettling due to the prevalent use of veto in real world organizations.
We show that this result is tied to the setup of the Dessein model as opposed to being a fundamental relationship between delegation and approval.
Dessein adopts the CS model and assumes a fixed difference between the optimal scale of the agent versus the principal for all project types. This rules out a beneficial pooling region under veto. This region is likely to be present in may real world firms. He also assumes a uniform distribution.
Our key points
In our model, project type is project quality and types are ordered from lowest to highest.
The agent wants to operate each project at a higher scale than would the principal due to private benefits of spending.
A perfectly informed principal would like to shut down an interval of the lowest quality projects, but an agent with complete decision rights would want to operate these at positive scales.
Approval can have an advantage over delegation because it induces agents observing low quality projects to pool at zero scale, as opposed to operating their low quality projects at a positive scale and generating losses for the principal.
If the probability distribution on types places enough mass on such types, we show that veto can dominate delegation for all feasible divergence parameters.
Intuitively, veto is better when it results in the rejection of low quality projects which have relatively high likelihood.
Use the implicit definition of θi to write θ1 as a function of θ2. Then we optimize the principal’s welfare over θ2.
The main result
We provide a simple model of a principal and an agent in an organization where two alternative decision processes can be used: delegation and veto.
We show that for any feasible magnitude of the difference in preferences between the principal and the agent, there is no clear-cut ranking of delegation and veto.
Approval has the advantage of shutting down low quality projects which if operated will cause losses for the principal, whereas delegation will result in all of these projects being operated at a positive scale. Delegation has the advantage over veto because it avoids the problem of marginal types mimicking the behavior of acceptable types and, in doing so, scaling their proposals beyond their own private optima.
Approval is most useful when lower quality projects are relatively more likely to be undertaken by a manager with decision making authority.