This is the whole course in one sitting. Hold the through-line in your head: the moment you hand a task to an agent to act on your behalf, two problems open up at once — their interests diverge from yours, and you can’t fully see what they do or what they know. That gap is the principal–agent problem, and everything you’ve learned is an attempt to close it: moral hazard (hidden action) and adverse selection (hidden information), the screening and signalling that fight back, the agency costs you pay no matter what, and an alignment toolkit — incentive pay, monitoring, reputation — where every tool shrinks the gap and every tool backfires if you lean on it too hard. The deepest lesson is that no contract is perfect: tie pay to a noisy outcome and you dump risk on the agent; reward one measurable thing and the unmeasured things get dropped. And the model itself lies if you forget that agents are more than money-maximisers — that trust and intrinsic motivation are real, and can be crowded out by the very incentives meant to help. Take a breath. There is no going back once you commit.
How this exam works
This is a real exam, not a practice quiz. Questions appear one at a time. Once you submit an answer it is locked for good — there is no going back, no retry, and no restart. Your score stays hidden until the very end. A few questions ask you to select all that apply (read those carefully — partial credit is not a thing here). You need 70% to pass. Ready when you are.
In the principal–agent problem, who is the PRINCIPAL and who is the AGENT?
Select an answer to continue.
Big picture
The principal–agent problem, in one picture
- Principal–agent problem
- The setup
- A principal delegates to an agent who acts for them; the problem needs TWO bricks at once - misaligned interests AND hidden action/information. Either one alone is harmless; delegation chains stack the roles all the way down, and you cannot fix it by hiring nicer people
- Hidden action = moral hazard
- Arises AFTER the deal, when the agent acts unobserved and doesn't bear the full consequences - two flavours: too little effort (the insured driver who stops locking the car) and too much risk (the bailed-out bank playing heads-I-win-tails-you-lose)
- Hidden information = adverse selection
- Present BEFORE the deal, when the worst types are keenest to deal - Akerlof's market for lemons unravels, insurance death-spirals, and risky borrowers crowd in at high rates; fought back with screening (uninformed party makes types self-sort) and signalling (informed party spends to reveal, only if cheaper for the good type)
- The price of delegation
- Agency costs (Jensen-Meckling) = monitoring + bonding + residual loss; even the best contract leaves a residual loss because closing the last of the gap costs more than it saves, and contracts are always incomplete - no rule foresees every contingency
- The alignment toolkit (and its backfires)
- Incentive pay/equity (Goodhart gaming, reckless option-risk), monitoring (costly, incomplete, crowds out goodwill), reputation & repeated dealing (end-game effects), efficiency wages (costly) - and the risk-incentive trade-off: tying pay to a noisy measure dumps risk on a risk-averse agent, so the optimal contract is interior and flatter the noisier the signal
- Where the model lies
- Agents are more than money-maximisers - intrinsic motivation and trust are real and can be crowded out (the daycare fine that raised lateness); the multitasking problem means rewarding the measurable drops the unmeasured; beware the over-cynical low-trust trap that destroys working goodwill
- The setup
Key takeaways
You now hold the whole model. The principal–agent problem is the anatomy of delegation: the moment an agent acts on your behalf, their interests diverge from yours and you can’t fully see their action or their knowledge — and you need both bricks for the problem to bite. Hidden action is moral hazard, arising after the deal in two flavours — too little effort (the insured driver who stops locking up) and too much risk (the bailed-out bank’s “heads I win, tails you lose”). Hidden information is adverse selection, present before the deal, where the worst types are keenest to deal: Akerlof’s market for lemons unravels, insurance death-spirals, and risky borrowers crowd in at high rates. Two weapons fight back — screening (the uninformed party builds a menu or test so types self-sort) and signalling (the informed party spends on a costly action that separates only if it’s cheaper for the good type). Delegation always has a price: agency costs = monitoring + bonding + residual loss (Jensen–Meckling), and that residual loss never goes to zero because contracts are incomplete and the last slice of alignment costs more than it’s worth. The alignment toolkit — incentive pay, monitoring, reputation, efficiency wages — shrinks the gap, but every tool backfires if over-used (gaming, reckless risk, crowded-out trust, end-game effects, cost). And the deepest constraint is the risk–incentive trade-off: tie pay to a noisy measure and you dump risk on a risk-averse agent, so the optimal contract stays interior and flattens as noise or risk-aversion rises. Finally, the model lies if you forget agents are more than money-maximisers: intrinsic motivation and trust are real and can be crowded out (the daycare fine that raised lateness), the multitasking problem means rewarding the measurable buries the unmeasured, and the over-cynical low-trust trap can destroy the very goodwill your relationship ran on. Never ask “how do I control this person?” Ask: what makes acting in my interest also their best response — without smothering the trust that was already doing the work?