In praise of principal-agent problems

When principals contract with independent agents who know things they don’t, there are costs. Contrary to common connotation, this is sometimes good. For example, we can view the failures outlined in Seeing like a state as warnings about incautious attempts to ignore or eliminate the principal–agent problem. Furthermore, we can view the principal–agent problem as a fundamental constraint on autocrats that limits them when compared to inclusive societies.

The concept

Briefly, principal–agent problems occur when a principal enlists an agent to act on their behalf, but the agent has private information and independent interests. This scenario means that the principal incurs agency costs when compared to simply acting themselves.

Examples include brokers front running, doctors prescribing medications based on their relationships with pharmaceutical companies, and taxis taking overlong routes.

Principal-agent possibilities

As these examples suggest, principal–agent problems are (I think) usually viewed as a bad thing. And they often are. But they can also be a refuge. To see this, we just have to flip our sympathies. The common examples encourage us to sympathize with the principal and see the agent as abusing their power.

Examples from the flipped perspective include:

If your manager at work is a petty tyrant, you surely appreciate and exploit whatever gaps in their knowledge you can find to work in ways you believe more effective or just to preserve your psychological health.
Overbearing parents
Some parents seem to think of their children as agents in the project of aggrandizing themselves. Children in this situation may justly find that moving out–an increase in agency costs—represents a gift.

The common theme here is that the principals are people of power misusing it and their ignorances open up little pockets of possibility for the less powerful.

Seeing like a state

We can put this perspective to more grandiose ends. For example, we can view the story of (Scott 1998) as the story of high modernists trying to reduce agency costs. In weak, early states, elite principals had limited capacity to bend the populace (agents) to their will. There simply wasn’t enough information for elites to make any but the most basic demands. It’s hard to conscript or tax citizens when you don’t know their names or even how many there are. From this principal–agent perspective, the celebrated ‘metis’ (local knowledge) is precisely the information asymmetry that justifies granting agents (some) freedom of action.

In other words, metis means that elite principals should delegate to agents because they don’t know everything. Principals respond to this fact by either trying to limit metis (to reduce agency costs) and make the people legible or by ignoring metis—to sometimes disastrous effect.

Why nations fail

(Acemoglu, Johnson, and Robinson 2005) talks about problems that arise in stratified societies due to commitment problems. I think we can also think of principal–agent problems as constraining autocracies (and other stratified, non-inclusive societies). A society in which there were no information asymmetries between autocrat and citizens would be the most perfect totalitarianism. Fortunately, no autocrat has yet achieved this and so autocrats must rely on other agents. These agents are assigned tasks with some latitude because the autocrat doesn’t know exactly how a task should be performed and task performance is evaluated with some latitude because the autocrat can’t know precisely how a task was executed. These twin latitudes create room for subversion or otherwise allow independent human expression rather than providing the autocrat with a nation of perfect puppets.

Just like in (Acemoglu, Johnson, and Robinson 2005), we can use this structural feature of stratified societies to explain the benefits of inclusive institutions. Principal–agent relationships will always be limited and brittle because the principal can only specify finite requirements. The success we see in principal–agent relationships often relies on some good faith on the part of the agent. But autocracies should often expect the principal–agent relationship to be adversarial. Under these conditions, contracts and explicit requirements are often woefully inadequate. On the other hand, inclusive institutions give everybody an independent stake in success—thereby converting agents into fellow principals and providing robust incentive alignment.

Finally, I’ll note that as mass surveillance improves, this advantage for inclusivity weakens. As mentioned previously, principal–agent problems only exist in the face of information asymmetries. To the extent that surveillance gives principals more information about agents and their actions, agency costs are reduced and autocracies become more effective. I’m certainly not the first to point out how mass surveillance can enable further domination, but I think this is a somewhat interesting angle on that observation.

Acemoglu, Daron, Simon Johnson, and James A Robinson. 2005. “Institutions as a Fundamental Cause of Long-Run Growth.” Handbook of Economic Growth 1. Elsevier: 385–472.

Scott, James C. 1998. Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press.