Moral Hazard

Why it matters

People take more risk when someone else is on the hook for the downside — and that’s not a moral failing, it’s the rational move, which is exactly why you fix the structure and not the person.

For example: a company tells its sales reps to expense any client dinner, no cap, fully reimbursed. Spending explodes — not because the reps turned dishonest overnight, but because the part of the decision that used to sting (the bill) now lands on someone else. Take away the consequence and you’ve quietly changed what the smart move is. Every rep is being perfectly reasonable. The reimbursement rule did the damage.

What it reveals. That riskier behavior isn’t a lapse of character but a predictable response to who carries the cost of failure — and that it will appear in anyone you put in that seat, not just the people you’d blame.
How it changes the read. You stop asking “why is this person being so reckless?” and start asking “who eats the loss if this goes wrong?” Find the gap between who decides and who pays, and the behavior explains itself.
When to foreground it. Any time a safety net went in — insurance, a guarantee, a bailout, an unlimited budget, someone else’s money — and behavior turned riskier afterward; or whenever authority over a decision is split from accountability for its outcome.
What you’d miss without it. That the fix isn’t a sterner talking-to or a better-hearted hire — it’s putting the decision-maker’s own skin back in the game, so prudence becomes their best move rather than a favor you’re asking for.
Where it misleads. Not every risky bet is moral hazard — some risk is just the right call for the actual odds. And it’s not the same as the wrong people signing up in the first place; that’s its sibling, adverse selection.

Realtime examples

See real, dated analyses where this pattern shaped the read on the news → Moral hazard on Main Street Independent

How to invoke it in Ora

You’re looking at behavior that turned riskier — an expense line that exploded, bets that got more aggressive, corners that started getting cut — and you suspect it’s because the person making the call no longer carries the cost of it going wrong. You want to know why, and how to put the consequence back.

Describe who makes the risky decision, who absorbs the loss if it fails, and the safety net that came between them, then ask:

“Moral hazard: once we fully reimbursed reps for any client dinner with no cap, spending exploded and we cannot audit each meal. Analyze the hidden-action problem and how to redesign the incentives.”

Ora names who is shielded from the downside, traces how that shield rationally pushes behavior toward more risk, classifies it as moral hazard rather than its pre-contract sibling, and works out the skin-in-the-game design that would make prudence the decision-maker’s own best move.

One thing to know: the words moral hazard are what route you here. A plain version — “our reps’ dinner spending is out of control, what do we do?” — gets a clarifying question back instead, because the bare situation could be a dozen different problems; naming the mechanism is what tells Ora which one you mean. Moral hazard, hidden action, skin in the game, principal-agent, or incentive design are the words that point it the right way.

Describe what the decision-maker can do once they’re shielded that the party bearing the cost can’t easily see or police — that post-contract hidden action is the whole mechanism. If instead the problem is that the riskier types were the ones who signed up to begin with, you’re looking at adverse selection, and Ora will say so.

One thing Ora won’t do: tell you to rip the safety net out. Insurance, guarantees, and backstops often exist for good reasons — spreading risk, preventing a collapse — so Ora calibrates the skin in the game rather than recommending you abolish the protection.

How it works

Picture two people with the exact same car, the exact same commute, parked on the exact same street. One has no insurance. The other is fully covered — every dent, every theft, every loss reimbursed down to the last dollar, no deductible.

Watch how they park. The uninsured one circles the block for the well-lit spot, noses in carefully, checks the mirror twice, maybe springs for the garage when the neighborhood looks rough. The fully-covered one pulls into the first space they see and walks away. Why wouldn’t they? A ding in the door is the insurer’s problem now, not theirs. Over a year, they stop replacing the smoke-detector battery, stop locking the steering wheel, stop sweating the things they used to sweat — because the policy pays out either way, and caution costs effort for a downside they no longer feel.

Neither of these people is a bad person. The covered driver isn’t trying to cheat anyone; they’re not even thinking about it. They’ve simply, and sensibly, stopped paying a cost that someone else now absorbs. Hand anyone that policy and the same thing happens. That is the whole point, and it is the part people get wrong: this is not about character.

That pattern has a name — moral hazard — and it’s a worse name than it deserves, because the word “moral” makes it sound like a failure of virtue. It isn’t. It’s a feature of structure. The economist Kenneth Arrow gave it its modern shape in 1963, studying why people with health insurance use more medical care than people without it. His answer wasn’t that the insured were greedy. It was that insurance, by design, breaks the link between a decision and its cost — and once that link is broken, any reasonable person leans toward the cheaper-feeling, riskier choice, because for them it genuinely is cheaper.

Here’s the engine, stripped to its gears. Every risky decision is a weighing: the upside if it works against the downside if it doesn’t. But you only weigh the part of the downside that lands on you. Insulate someone from the downside — with insurance, a bailout, a blank check, a job they can’t be fired from, an “undo” button — and you’ve lightened one side of their scale without touching the other. So the scale tips toward risk. Not because they’ve changed. Because the math has. Take the CEO whose contract guarantees a golden parachute: if the big gamble pays off they get rich, and if it craters they walk away whole. The aggressive bet isn’t recklessness on their part — given that contract, it’s the rational play. The contract built the recklessness in.

And that is exactly why you can’t fix moral hazard by firing the reckless person and hiring a careful one — the careful one, in the same seat, eventually parks just as badly. You fix it by putting their skin back in the game: a deductible so the driver feels the dent, a co-pay, a clawback that takes the bonus back if the bet later blows up, equity that vests over years so the CEO can’t cash out before the consequences arrive, a requirement that they invest their own money alongside everyone else’s. Each one stitches the cost back onto the person making the call. The aim isn’t to punish — it’s to make prudence their best move again, so you don’t have to ask for it. Separate a decision from its downside and any rational actor takes more risk. Reconnect them, and the recklessness simply evaporates — because it was never really about the person.

Framework & implementation

This section uses Ora’s own terms for the parts of an analysis, so that if you open the actual mode and lens files they line up. Each is glossed in plain language on first use.

Pipeline execution

Moral hazard is one of the mental models in Mechanism and Incentive Analysis’s ANALYTICAL PERSPECTIVES block, listed under “always loaded” — so it is active on every run of that mode, whether or not the prompt names it. (Mechanism and Incentive Analysis is the information-and-incentive sibling of Strategic Interaction: where that mode reads a game of observable moves, this one handles the case where what each party privately knows or privately does is the crux.) It runs at Gear 4, Ora’s most thorough setting: a Depth analyst and a Breadth analyst read the situation independently, each critiques the other’s reading, both revise under that critique, and a consolidator merges what survives. The lens threads through those stages like this.

Detection. The lens engages on the cases in its Detection Signals — a safety net, guarantee, or insurance went in and behavior shifted toward more risk afterward; decision-makers spending resources they do not own (other people’s money, time, reputation); accountability for outcomes split off from authority over the decision; a known system backstop the actors can lean on; asymmetric pay that rewards the upside but doesn’t bite on failure. The precondition is the mode’s first Critical Question (CQ1, asymmetry named): one party can take an action whose downside falls on another, and that action can’t be fully seen or policed by the party who pays.

The Depth and Breadth analysts. Two models read the situation in parallel. The Depth analyst commits to one reading and defends it — it names who makes the risky decision and who bears the cost if it fails, then runs the lens’s Application Steps: check whether the decision-maker’s downside actually matches the risk they’re taking, and look for the behavior turning riskier after a safety net was introduced or strengthened. Its load-bearing job is the mode’s CQ2 (selection vs. hazard): classify this as moral hazard — hidden action, the behavior changing after contracting — and not its pre-contract sibling, adverse selection, which sorts risk types into the system before anyone signs. The Breadth analyst works the same situation at the same time, scanning the mode’s other information-and-incentive failure modes — adverse selection, the principal-agent split, signaling, the commons dynamic — and ruling each in or out rather than assuming this one. Neither analyst sees the other’s work.

Cross-adversarial evaluation. Each analyst’s reading is handed to the other to critique against the mode’s criteria. The lens’s signature failures are caught here, keyed to its Critical Questions: blaming the actor’s character instead of the structure (character-attribution substitution — the tell is a fix that amounts to firing or naming-and-shaming, when the same structure would produce the same behavior in anyone), reading pre-contract sorting as post-contract behavior change (adverse-selection conflation — the mode’s CQ2, the one distinction it will not let blur), and calling ordinary risk-taking moral hazard when it’s really the appropriate response to the actor’s actual odds (the mode’s CQ1 on whether the risk is genuinely elevated above what the actor’s profile would predict). The evaluator also presses the mode’s CQ3, that any proposed fix carry both design constraints.

Revision and claim-check. The reviser addresses the fixes. Where the reading rests on a factual claim — that spending actually rose after the policy changed, the real shape of a compensation package, who is genuinely shielded — that claim is marked a flagged claim and sent to a web-search tool; it has to resolve against outside sources before the revised draft moves forward.

Consolidation and output. The consolidator merges the two revised readings into one information-and-incentive corpus, and the formatter places it into the mode’s set sections. Who makes the decision and who bears the downside lands in Parties and the asymmetry. The classification — moral hazard, hidden action, post-contract — lands in Selection vs. hazard, held strictly apart from adverse selection. The shift toward risk once the downside is shielded lands in The distortion. The skin-in-the-game devices (co-pays, deductibles, clawbacks, equity vesting, personal co-investment) are grounded in Named mechanisms in play (the mode’s CQ4 forbids name-dropping — each must be shown operating, and enforceable, since an unenforced clawback is no countermeasure). When the prompt asks for a fix, the design lands in Mechanism (for a design), which must carry both the participation constraint (will the party still take the deal once real downside is attached? — over-loaded skin in the game and the CEO simply walks) and the incentive-compatibility constraint (does bearing that downside make prudent behavior their own best response?) — a design that satisfies only one is reshaped. The conclusion lands in Read, with the mode’s CQ5 posture marker stating plainly whether it is explaining a failure or proposing a fix.

What the analysis will not assert. It reports the structural pull toward risk and what would re-couple decision to consequence. It does not treat the behavior as a morality failure — moral hazard is structural, and the mode’s hyperrationality caution cuts both ways here: the same incentive produces the same behavior in any rational actor, so the fix is the structure, never the person’s character. Nor does it recommend tearing out the safety net — insurance, guarantees, and backstops often serve legitimate ends (allocating risk to those who can bear it, preventing systemic collapse), so the lens’s safety-net-abolition caution keeps the task at calibrating skin in the game, not eliminating protection. And it holds moral hazard strictly apart from adverse selection, because the two demand different fixes.

Origin and evidence

The modern treatment is Kenneth Arrow’s, from his 1963 American Economic Review paper “Uncertainty and the Welfare Economics of Medical Care,” which introduced moral hazard into economics by showing that health insurance, by absorbing the cost of care, rationally raises how much care the insured consume — not through any dishonesty, but through the structure of the coverage. Mark Pauly’s 1968 AER comment sharpened the point against the moralizing reading: the extra consumption is the predictable response of a rational actor to a lowered price, not “moral perfidy,” which reframed moral hazard as an incentive problem rather than an ethics problem and effectively launched the field. Bengt Holmström’s 1979 Bell Journal of Economics paper “Moral Hazard and Observability” gave it its formal principal-agent spine, deriving when tying a party’s pay to observable signals of their hidden action can improve on paying for outcomes alone — the analytic basis for every skin-in-the-game contract that follows. The structural-countermeasure framing also reaches a general audience through Nassim Nicholas Taleb’s Skin in the Game (2018), which argues that those who make decisions should bear a share of the downside as an ethical and systemic principle.

Applications and common uses

Moral hazard is a workhorse diagnosis wherever authority over a risky decision is split from accountability for its cost — used both to explain why behavior turned reckless and to design the skin in the game that re-couples the two.

Insurance. The originating case: coverage that absorbs the loss dampens the incentive to prevent it, so insurers re-attach a share of the downside through deductibles, co-pays, and coinsurance — enough to restore care without gutting the protection people bought the policy for.
Banking and bailouts. A firm confident it will be rescued if a bet goes bad — “too big to fail” — rationally takes bigger bets, since the taxpayer holds the tail risk. The structural answers are capital requirements, living wills, and clawbacks on bonuses, each forcing the decision-makers to carry more of their own downside.
Executive compensation and corporate governance. Asymmetric pay — large upside, no real downside — invites aggressive bets with the company’s resources. Boards re-couple decision to consequence with long-vesting equity, clawback clauses, and required personal co-investment, the textbook fix for the golden-parachute problem.
Organizations and “other people’s money.” Whenever someone spends a budget, time, or reputation that isn’t theirs and can’t be closely watched, spending drifts toward the wasteful; the response is ownership of a P&L, approval thresholds, and audit — accountability placed back where the authority sits.
Mechanism and contract design. Run in reverse, the lens is a design brief: build the deductible schedule, the vesting curve, or the performance-tied pay whose terms make the prudent action the decision-maker’s own best move — designing the incentive so it satisfies both the participation and incentive-compatibility constraints, rather than policing behavior after the fact.

In every case the diagnosis is the same: the decision has been split from its downside, that split makes the riskier choice the rational one, and the fix is to reconnect them — not to wish for more virtuous actors in a structure that rewards the opposite.

Failure modes and when not to use it

The lens’s characteristic ways of going wrong are catalogued in its Common Failure Modes:

Safety-net abolition. Using the diagnosis to argue for removing all protection. The tell is a recommendation of “no safety net” rather than “calibrated skin in the game.” Insurance, guarantees, and backstops often serve real purposes; the task is to weigh those against the moral-hazard cost and calibrate, not eliminate.
Character-attribution substitution. Treating the elevated risk-taking as a moral failure of the actor rather than a structural response. The tell is a fix that amounts to firing or naming-and-shaming. The same structure produces the same behavior in any actor — so change the structure, not the person.
Unenforceable countermeasure. Designing skin-in-the-game provisions that can’t actually be enforced. The tell is a clawback that’s never been exercised or a deductible that’s waived in practice. A countermeasure on paper is no countermeasure; auditability and pre-commitment are what make it bite.
Adverse-selection conflation. Reading pre-contract sorting as post-contract behavior change. The tell is that the elevated risk comes from the type of actors who entered the system, not from a change in how existing actors behave — that’s adverse selection, a different mechanism with different fixes. Keep the two apart.
Creep blindness. Missing the gradual rise in risk-taking that follows a safety net going in. The tell is a behavioral shift that accumulated quietly over years and wasn’t noticed until something broke. The correction is to monitor risk-taking actively after any change to the safety net, because moral hazard compounds over time.

When not to reach for it. When the decision-maker fully bears their own downside, there is no asymmetry to drive the behavior and the lens predicts a recklessness that won’t appear. When the risk-taking is simply the appropriate response to the actor’s real odds — not an artifact of an externalized downside — calling it moral hazard misreads sound judgment as a structural defect. And when the elevated risk traces to which actors entered the system rather than to changed behavior among those already in it, the mechanism is adverse selection — reach for that sibling instead, because re-attaching downside to current actors won’t fix a pool that was mis-sorted on the way in.

Mechanism and Incentive Analysis — the analysis that hosts this lens; reads situations where hidden information and incentive structure drive the outcome, and designs the rules that fix them.
Adverse Selection — the pre-contract sibling: the riskier types sort into the deal before anyone signs. Moral hazard changes behavior after the deal; adverse selection sorts the pool before it.
Principal-Agent Problem — the broader frame of one party acting for another under hidden action and misaligned incentives, of which moral hazard is the post-contract, hidden-action face.
Tragedy of the Commons — the adjacent dynamic: a shared resource is over-used because each actor doesn’t bear the full cost of depleting it — the same split between decision and downside, spread across a group.