principal-agent problem

I think we should develop this a little more.

GPT5.icon

Principal = delegator, agent = assignee. The problem is that the agent does not take the action desired by the delegator due to the misalignment of the interests of the two parties and the asymmetry of information (not being able to see the actions and capabilities of the other party).

Main type

Decomposition of agency costs (Jensen & Meckling-like arrangement)

Monitoring costs + pledge/collateral costs (bonding) + residual losses.

**Digging into the decomposition of proxy costs (Jensen & Meckling, 1976)**. The gist is.

Agency Cost = Monitoring + Bonding + Residual Loss.

1. contents of the three elements

Monitoring costs (expenditures on the part of the principal)

Cost to "visualize/constraint" agent behavior and outcomes.

Examples: external audit and internal controls, board of directors/independent external take, KPI/log measurement, compliance/risk management, quarterly disclosure, auditor fees, coaching/evaluation interviews, etc.

Pledge/bonding costs (on the agent's part)

The expected cost an agent incurs in assuring itself that it will "do no harm/keep its word."

Examples: deferred compensation linked to performance (lost if not achieved), equity investment (shareholding or co-investment), performance bond (surety bond), clawback clause, opportunity loss due to non-compete or confidentiality, penalty clause, self-funded professional liability insurance, etc.

The "grant" of stock options itself is not an immediate cost, but bonding is a "mechanism that will hurt you if you fail to protect it," such as losing/forfeiting restricted stock if you fail to achieve it.

Residual losses (efficiency losses that remain after both parties have incurred costs)

The diminution in value that still remains compared to the "ideal" (first best), in which the interests are perfectly aligned and fully observable.

For example, the NPV may be slightly lower due to management's "preference for scale," a little less visible effort, short-term indicator optimization may slightly damage the long-term value, and so on.

2. intuitive design principles (how much to spend?)

The objective is to minimize the sum of M (monitoring) + B (bonding) + RL (residual loss).

Increasing monitoring will lower RL but increase M. If bonding is increased, RL will go down but B will increase, which will also cause recruitment difficulties and higher opportunity cost of human resources.

The basic principle is to stop at marginal benefit = marginal cost.

6) Common misconceptions

Maximizing surveillance is all that's needed" → Excessive costs, low morale, inhibits creativity.

Maximizing the outcome linkage is all that is necessary."→ Risk shifting, short-term, and disregard for unmeasured value.

Bonding = only penalties" → applies to all "self-inflicted damage mechanisms" such as interest sharing, self-investment, clawbacks, etc.

Summary:

The proxy cost is a design problem that aims for the minimum sum of "the cost of watching (M)", "the cost of binding oneself (B)", and "the loss that still remains (RL)". The practical trick is to stop signaling marginal benefit = marginal cost, in the order of visualization, interest sharing, and minimum necessary restraint.

---