Third Party Moral Hazard — Peter Siegelman

Peter Siegelman · 2022 · 1:06:56 · ▶ Watch

Summary

Central question. We have written thousands of articles on how insurance changes the behaviour of the policyholder (first-party moral hazard). What does insurance do to the behaviour of people who are not in the contract?

Thesis. Insurance creates a pool of money, and outsiders can tap it. Siegelman defines third-party moral hazard as “the effects of insurance on the loss or claims increasing behavior of people who are not party to the insurance contract”. The authors’ claim is modest and deliberately so: “this is a thing and there are reasons to take it seriously.” That second half is my paraphrase of; the exact words are in the transcript as “our thesis is just very simple it’s this is a thing and there are reasons to take it seriously” across.

The structure of the argument.

1. Definition and picture. A tangle of illegal taps on a power line. The insurance pipeline is the power line; third parties tap it.
2. Examples. A staged motorcycle crash in the UK, the “Florida shuffle” in addiction treatment, kidnap and cyber ransom, health-care supply (MRIs, copay coupons), no-fault billing rings in New York, D&O claims.
3. Why it matters. It lowers insurance’s net benefit, as ordinary moral hazard does. It may be as big as first-party moral hazard. The usual insurer tools (deductibles, underwriting, loss control) do not reach it because there is no contract with the loss causer.
4. Mechanisms. Deep pockets; an impersonal payer (it feels different to take money from an insurer than from a person); insurance crowding out monitoring.
5. Remedies, tentatively. Market design (the Lloyd’s kidnap market), the D&O “duty to reimburse” as a deterrent to strike suits, subsidising precautions, and a qui tam bounty system for private fraud-hunters. The state has to help (“government as governance”).

How it ends. The scholarly bottom line is that insurance should be studied “in a general equilibrium context”: look up and out from the contract to the ripple effects on everyone’s incentives. In the Q&A the main objections are (a) subrogation, (b) how much of this is fraud versus mere carelessness, (c) the moral difference between a hospital raising prices and the Florida shuffle, and (d) the political use of the term “moral hazard”. Siegelman concedes that he has “no grounding in moral philosophy” and cannot yet say which third-party moral hazard is bad.

Stance to note. He is candid that the evidence is anecdotal: it is “not of the causally robust social science kind”. He wants the idea named and put “on people’s radar screen”, not proven.

Map — talking points

#	Point
1	Introduction by the host; announcement of next week’s flood-insurance talk
2	Definition: insurance effects on loss- or claims-increasing behaviour of non-parties. Not a pollution-style externality: it opens an opportunity
3	Picture: tapping a power line. Caveat: evidence is anecdotes, not causal social science
4	Example 1: staged motorcycle crash (UK compulsory cover)
5	Example 2: the Florida shuffle (drug rehab centres, patient brokers, relapse as a feature)
6	Why care: lowers net benefits of insurance; Arrow; magnitude; primary, secondary, adverse-selection losses
7	Extra-contractual: no privity. Tools of insurance-as-governance fail; need the police; “government as governance”
8	Loss control as displacement: the kidnap wall; EPLI training “bulletproofs the policyholder”
9	Under-litigation caveat; malpractice study; third-party moral hazard is not the same as fraud
10	Kidnap: ransoms, Italy and Colombia, the Lloyd’s soft cartel; cyber ransom
11	Health care: “overcharged”; Nyman’s good moral hazard; wallet biopsy; MRIs; Arrow on physicians’ ethics
12	Copay coupons
13	Liability insurance: no-fault billing ring in New York; D&O evidence is weaker
14	Mechanisms: deep pockets; Baker’s “blood money”; crowding out monitoring (car repair, too-big-to-fail creditors)
15	Remedies: Lloyd’s market design, D&O duty to reimburse as pre-commitment, dashcam subsidies, qui tam bounties
16	Bottom line: general equilibrium view of insurance
17	Q&A, Maher: subrogation as an equitable, non-contract remedy; why it is limited here
18	Q&A, Maher: fraud versus extra negligence; Siegelman admits a biased sample
19	Q&A, Baker: gradations; “can’t be bothered” moral hazard; kickback anecdote; naming effects; ideology of the term
20	Q&A, host: how to tell good from bad; intent; copay coupons; “no grounding in moral philosophy”
21	Q&A, Yass: Florida assignment of benefits (roofing, auto glass)
22	Q&A, Maher: ERISA first-dollar reimbursement kills victims’ incentive to sue negligent third parties

Putting it together

One-paragraph mental model. Every insurance contract is a pipe that moves money from a pool to claimants. First-party moral hazard asks what happens to the person at the tap. Third-party moral hazard asks what happens to everyone standing near the pipe: people who stage crashes, doctors who price to coverage, drug makers who cancel the insurer’s price signal, creditors who stop watching, and tortfeasors who expect somebody else to pay. The pipe creates a prize; the prize draws effort; some of that effort is waste and some is crime, and some is just ordinary pricing. The insurer’s tools (underwriting, deductibles, claims handling) work through the contract, so they miss people outside it. The fixes therefore sit outside the contract too: concentrated and watchful buyers (Lloyd’s), precommitments (D&O), enforcement by the state or by paid private hunters (bounties), and rules about who may touch the pool. Meanwhile, because claiming is often too low, a crude attack on claims can do harm; the hard step is to tell the predatory from the valid.

Open tensions.

• Breadth versus bite. A concept wide enough to cover staged crashes and routine price-setting may not guide any single remedy.
• Insurer incentives. Giving insurers more power to deny claims runs into the “dark side” argument that they do not always want to cut losses.
• Evidence. The cases are vivid and the measurements are absent. Siegelman says so; the field has to build designs that can separate the effect of the pool from that of other causes.
• Naming. A name helps people see the pattern and can also become a political weapon.
• Displacement. Every local fix may move the activity elsewhere.

Reading path.

• Start: the paper itself (journal page; a draft is on Stanford’s site), then Arrow’s 1963 paper.
• Then: Baker’s work on the genealogy of moral hazard and on insurance as governance; Shortland’s Kidnap; the Florida shuffle article in the New Yorker.
• Deeper: Ben-Shahar and Logue on outsourcing regulation; Avraham and Porat on the dark side of insurance; the criminology of opportunity (Cohen and Felson; Clarke); and the economics of rent-seeking (Tullock, Krueger, Baumol).

Part 1 — Before you read

Key terms

• Moral hazard. Watch out: three senses. (1) Modern economics (Arrow, Pauly, Shavell): insured people take less care, or use more care-related goods, because someone else pays. The change is in behaviour caused by the contract. (2) Older insurance-trade usage: the moral character of the applicant (a dishonest or careless person is a moral hazard), as against physical hazard (a wooden building). Baker’s historical work† traces how the word moved from sense 2 to sense 1. (3) Popular and political usage: any bailout that “encourages risk-taking”. Siegelman’s definition keeps sense 1 but moves the actor.
• First-party moral hazard. The policyholder’s own reduced care. Fixed, in theory, by deductibles, coinsurance, experience rating, exclusions, monitoring.
• Third-party moral hazard. Insurance changes what non-parties do: they create losses, inflate claims or drop their guard, because an insured pool of money exists. Siegelman’s examples show three sub-types: (a) predation: staging or inflating claims to reach the money; (b) price capture: hospitals, drug makers, repair shops charging more because an insurer pays; (c) crowding out of monitoring: creditors, repair customers or victims stop policing because the insurer is behind the actor.
• Externality. A cost or benefit that one party’s action imposes on others without a price. Siegelman says his case differs from pollution: the third party’s actions cause the losses, and what insurance did was to open “scope for loss creation”. In classic terms, insurance creates the incentive, not the harm itself.
• Privity. The legal link between the parties to a contract. Absence of privity means an insurer cannot write terms with the loss causer.
• Subrogation. The insurer, having paid its insured, steps into the insured’s shoes to sue whoever caused the loss. Maher notes it began as an equitable doctrine, not a contract term.
• Insurance fraud. Lying to obtain payment. Siegelman says it is both too wide and too narrow a category for his topic: arson by the policyholder is fraud but not third-party moral hazard; kidnap and copay coupons are third-party moral hazard but not insurance fraud.
• Duty to reimburse (D&O). A policy under which the insurer simply pays defence costs as chosen by the policyholder, instead of controlling the defence. Siegelman argues this can deter strike suits.
• Qui tam. Latin shorthand for a private person suing on behalf of the state or another injured party and taking a share. The False Claims Act model†.

People named (a reading list)

Person	Why they come up	Start with
Peter Siegelman and Gideon Parchomovsky	The paper	”Third-Party Moral Hazard and the Problem of Insurance Externalities”, Journal of Legal Studies vol. 51 (journal page); an earlier draft is on Stanford’s site
Kenneth Arrow	Famous point on moral hazard; also his remark that physician ethics should limit overuse	”Uncertainty and the Welfare Economics of Medical Care” (1963)†
Tom Baker	Insurance as governance; “blood money”; his own cyber/kidnap work with Anja Shortland; on the call	”On the Genealogy of Moral Hazard” (1996); “Blood Money, New Money…” (2001); with Griffith, Ensuring Corporate Misconduct (2010) (all from memory)†
Anja Shortland	How the Lloyd’s kidnap market governs itself	Kidnap: Inside the Ransom Business (OUP, 2019) (search summary; source not confirmed)
Kyle Logue	Insurance and cyber moral hazard	Ben-Shahar and Logue, “Outsourcing Regulation: How Insurance Reduces Moral Hazard” (2012)†
Shauhin Talesh	EPLI insurers push compliance training	His work on insurers and compliance†
David Hyman and Charles Silver	Overcharged (health costs come from insurance)	Overcharged: Why Americans Pay Too Much for Health Care (2018)†
John Nyman	”Good” moral hazard: access that insurance makes possible	The Theory of Demand for Health Insurance (2003)†
Ronen Avraham and Ariel Porat	”The dark side of insurance”: insurers’ incentives not to stop losses	EconPapers entry (Review of Law & Economics, 2023)
Brendan Maher	Subrogation, ERISA	His insurance-law scholarship†

Siegelman’s own frame

• He calls it “kind of like a one idea uh paper”. The ambition is to name a phenomenon and show it is plausibly large, not to estimate it.
• He treats insurance law as more than contract: “insurance is just not pure contract”. That is the move that connects the paper to the “insurance as governance” literature.
• He is open about the limits: no causal evidence, a biased sample of vivid cases, and no theory of which cases are bad.

Current debate

• Is the idea new, and is it measured? The paper is in a peer-reviewed law-and-economics journal (journal page). A search summary says that an earlier Stanford-hosted draft was titled “Moral Hazard Due to Parties Other Than Insureds” (search summary; source not confirmed). I found no study that measures third-party moral hazard directly, which is also Siegelman’s point. That absence could change; I did not search the 2023-26 empirical literature in depth.
• The “dark side of insurance” line. Avraham and Porat’s article argues from the insurer’s side that insurers may have incentives to keep risks alive and may even encourage harmful behaviour by insureds or third parties to boost demand for cover (EconPapers entry, Review of Law & Economics 2023, vol. 19; the summary of its argument is from a search summary, only the title and venue are confirmed). Siegelman cites an early version. The two papers are complementary: one looks at who benefits when losses persist (insurers), the other at outsiders who manufacture claims.
• Cyber ransom is the live case. A TechTarget piece reports that critics call ransom reimbursement a troubling practice; it quotes a former US deputy national security adviser, Anne Neuberger (TechTarget). A search summary adds that New York’s financial regulator flagged moral hazard in cyber cover, that the insurers’ trade group opposes any ban, that AXA stopped selling ransomware cover in France and that the US Treasury’s sanctions advisory also reaches insurers (search summary; source not confirmed). I did not find a source confirming whether any US federal ban on reimbursing ransoms has been enacted, so I make no claim on that.
• India (my inference, not the speaker’s). The structure fits Indian motor third-party cover, which is compulsory. News sites report that the Supreme Court has ordered state-level special investigation teams on fake motor accident claims; the SCC Online headline says “SC Orders Pan-India SIT Probe” (headline and the term “SIT” confirmed; the details of the order, including the dates, are from search summaries and not confirmed). That is third-party moral hazard in nearly pure form: the third party is the claimant, and the compulsory pool is the power line. For public health cover such as Ayushman Bharat, hospitals billing a payer is the “price capture” pattern. I did not find a source in my search for fraud in that scheme, so treat any claim about it as† For AI liability cover, the same logic is a question, not an answer: if insurers underwrite AI harms, who gets to claim, and what claims become worth manufacturing?

The ideas

📚 Everything in this part is added background, not the speakers’ claims. Siegelman names only a handful of thinkers (Arrow, Baker, Shortland, Nyman, Hyman and Silver, Talesh, Logue). The paper sits at the meeting point of at least eight traditions. Names, dates and titles below are†, unless a link is given. I have tried to keep to what I am sure of and to say “I believe” where I am not. Read the section once, slowly, then use the ideas as the places where each tradition bites.

How the traditions meet

                      THIRD-PARTY MORAL HAZARD
                (insurance changes what NON-PARTIES do)
                                  |
   ---------------------------------------------------------------
   |            |             |            |            |         |
 Trad 1       Trad 2        Trad 3       Trad 4       Trad 5    Trad 6-8
 Moral        Externality   Insurance    Tort,        Crime,    Domain cases:
 hazard       / rent-       as           claiming,    opportun- health supply,
 (Arrow,      seeking       governance   under-       ity,      safety-net
 Pauly,       (Pigou,       (Ericson,    claiming     displace- monitoring,
 Shavell)     Tullock,      Baker,       (Calabresi,  ment      ransom
              Baumol)       Logue,       Galanter,    (Becker,  (Arrow,
                            Shortland)   Priest)      Cohen &   Kane,
                                                      Felson)   Schelling)
   |            |             |            |            |         |
   +---> SAME PARENT      <---+            |            |         |
         CONCEPT: tension    ---> WHY CONTRACT FAILS    |         |
         of risk spreading        (no privity)          |         |
         and risk creation             |                |         |
                                       +--> WHAT TO DO: state, market design,
                                            bounties; but "good" claims exist
                                            (Trad 4) and displacement limits
                                            loss control (Trad 5)

---

Where each school of thought is covered:

• The economics of insurance and moral hazard: idea 1
• Externalities, rent-seeking and unproductive entrepreneurship: idea 1
• Insurance as governance, and law’s old defences against third-party manipulation: idea 2
• Tort, deterrence and the economics of claiming: idea 3
• Crime, opportunity and displacement: idea 6
• Health-care economics: supplier-induced demand and the “price capture” side of insurance: idea 5
• Safety nets, bailouts and the erosion of market monitoring: idea 7
• Ransom, extortion and the credibility of non-payment: idea 4

1. What the concept is, and what it is not

In short. Siegelman defines third-party moral hazard as the effect of insurance on the loss-increasing behaviour of people who are not party to the insurance contract. The harm is not caused directly by the insured, and insurance opens scope for loss creation. His picture is people tapping a power line: where a conduit of money runs, someone will figure out how to tap it. It is not simply fraud, and he says his evidence is anecdote and not robust social science.

🎙 In the talk

• The definition: “the effects of insurance on the loss or claims increasing behavior of people who are not party to the insurance contract”.
• The contrast with pollution. Here the harm is not caused directly by the actor; what insurance does is to open “scope for loss creation”. Siegelman says there can also be gains from it, but he focuses on losses.
• The picture: people tapping a power line. His reading is that where a “conduit of money” runs, someone will figure out how to tap it.
• It is not fraud. Fraud is over-inclusive (policyholder arson), and some third-party moral hazard (kidnap, copay coupons) is not insurance fraud.
• Anecdote, not social science: “not of the causally robust social science kind”.

📚 Deeper theory

1. Naming a gap. Baker’s reaction is that “once you give something a name” you start to see a connection among cases. That is how the first-party concept grew too (Roots).
2. Opportunity versus harm. Roots supplies the framing: the pool is a rent and third parties are rent-seekers. idea 6 gives the criminologist’s version (target, offender, absent guardian).
3. Why not just call it fraud. Fraud is a legal wrong defined by deception. The concept here is defined by an economic effect regardless of legality: a hospital raising its price is lawful and still counts. That choice buys breadth but loses a clear line between bad and merely profit-seeking, which is the very problem the Q&A presses.

📚 Roots: The economics of insurance and moral hazard

The problem. People buy insurance to swap a risky loss for a certain premium. But insurance changes behaviour, and changed behaviour changes the loss. How much cover can a market offer before the cover undermines itself?

The point. The paper is built to mirror first-party moral hazard. Same parent concept (Arrow’s tension between “risk spreading and risk creation”), same welfare logic (net benefit of insurance goes down), same counting of costs (primary losses, secondary losses, adverse selection). The break is in the actor. In the textbook model the insured chooses care. Here a non-party chooses whether to create or inflate a claim.

Thinker chain.

• Nineteenth-century trade usage. “Moral hazard” first meant the dishonesty or carelessness of the applicant, as against “physical hazard”. The insurer’s worry was character: would this person burn the shop? Baker’s genealogy of the term (1996, from memory) traces how the word was later narrowed into the economist’s sense of incentive effects.
• Kenneth Arrow (1963), “Uncertainty and the Welfare Economics of Medical Care”. Medical care is a market where the usual assumptions break: the patient does not know what he needs, the doctor does, and insurance blunts price signals. Arrow uses “moral hazard” in the incentive sense and notes the tension between spreading risk and creating it. Siegelman quotes Arrow’s remedy for overuse: “rely on the professional ethics of physicians”.
• George Akerlof (1970), “The Market for Lemons”, opened the information-asymmetry programme. Rothschild and Stiglitz (1976) applied it to insurance: adverse selection (the risky self-select into cover) versus moral hazard (cover changes behaviour). Siegelman names “adverse election” as a way third-party effects feed back: higher premiums push out the good risks.
• Steven Shavell (1979), “On Moral Hazard and Insurance”, and Bengt Holmström (1979), “Moral Hazard and Observability”, gave the formal treatment. If the insurer could observe care it would write it into the contract. Because it cannot, it uses proxies: deductibles, coinsurance caps, experience rating, exclusions.
• John Nyman (1999, 2003) argued that much of the extra use caused by insurance is not waste: insurance is an income transfer that lets sick people buy care they could not otherwise afford. Siegelman uses this to say the doctor-visit story is not the one worth worrying about.

Core argument in more detail. Take the standard model. An insured person picks a level of care that costs her something and lowers the chance of loss. Without insurance she picks the level that balances her marginal cost against her own marginal benefit, so she internalises everything. With full insurance her marginal benefit falls to zero and she picks too little care. The policy response is partial cover: leave her some skin in the game. That gives the textbook trade-off, where the best contract trades risk-spreading against incentives.

Now add a third party. The third party’s payoff depends on the size of the insurer-funded pool he can reach. In the staged crash, the payoff is the claim against a compulsory policy; in the Florida shuffle, the rehab centre’s revenue per patient-month; for the repair shop, the markup it can charge an insured customer. A bigger pool raises the payoff to manipulating it, so the third party invests more. Crucially, the insurer cannot fix this by adjusting the insured’s cost-sharing. A higher deductible changes what the policyholder pays, but the staged-crash gang never sees it; their gain is the policy limit and the victim’s status as insured. This is why Siegelman asks “could you imagine in the motorcycle crash that a higher deductible or more careful underwriting would have made any difference”.

Where it is contested. Whether moral hazard is large at all is argued. The RAND results and Nyman’s point suggest much extra use is valuable. A second dispute is whether “moral hazard” is a descriptive term or a rhetorical one (Baker raises this). A third is whether a contract can hold the whole effect. Siegelman says no.

📚 Roots: Externalities, rent-seeking and unproductive entrepreneurship

The problem. Some activities make people rich without making the world richer. Where they happen, it matters whether the resources spent chasing the prize are a pure waste.

The point. Siegelman is explicit that third-party moral hazard is not the standard factory-smoke externality: it is “the creation of an opportunity to create losses”. Put in the older vocabulary, the insurance pool is a rent, and the staged-crash gang, the body brokers and the MRI-owning physician are rent-seekers drawn to it. His list of costs (primary losses, “secondary losses” in creating and detecting the scheme) is the Tullock/Posner point: the resources spent chasing a transfer are a social loss, even when the transfer itself is just a movement of money.

Thinker chain.

• Arthur Pigou (1920) named divergences between private and social cost; the tax remedy. Ronald Coase (1960) replied that with low transaction costs, bargaining fixes externalities, and that the law should look at who can avoid the harm cheaply.
• Jacob Viner (1931) and James Meade (1952) separated technological from pecuniary externalities. A pecuniary externality works through prices and is, in the textbook view, not itself a loss. A technological one changes the production possibilities.
• Gordon Tullock (1967) showed that the cost of monopoly and tariffs includes the resources spent to obtain them. Anne Krueger (1974) named “rent-seeking”. Richard Posner (1975) argued that those competing for a monopoly may dissipate the whole rent. Jagdish Bhagwati (1982) called the wider family “directly unproductive profit-seeking” activities.
• William Baumol (1990) distinguished productive, unproductive and destructive entrepreneurship: the same talent goes where the payoff structure sends it.

Core argument in more detail. Why does it matter that the externality is of opportunity, not of harm? Because the usual remedies differ. For smoke you tax or regulate the smoke. For an opportunity you can (a) shrink the pool, (b) change the rules that decide who gets paid, (c) raise the cost of reaching it, or (d) police the reaching. The bounty proposal is (d); the Lloyd’s practice of not paying the ransom itself is (a); a fraud investigation unit is (c). The Pigou–Coase frame also asks who is the cheapest avoider. In the staged-crash case, the cheapest avoider is the police or the road (cameras), not the insurer. That is the intuition behind “something bigger”.

Where others push back

• The concept may be too broad to be useful: it covers conduct from violent crime to routine pricing. Baker raises the gradations.

What they leave out

• Whether insured losses are more likely than uninsured ones in the same setting. Without a comparison group, one cannot say the pool caused the activity.

Check yourself

1. Why does Siegelman say the externality is not like smoke from a chimney?
2. Give one example that is third-party moral hazard but not insurance fraud, and one that is the reverse.
3. What kind of evidence does he say he has, and what kind is missing?

---

2. Why it matters, and why contracts cannot fix it

In short. Siegelman says third-party moral hazard lowers the net benefit of insurance, a tension between risk spreading and risk creation, though it does not make insurance bad. It can be as large as the first-party kind yet has had almost no attention. Because the loss causer has no contract with the insurer, it is largely extra-contractual. So insurance is not pure contract, and it may need police or government on top.

🎙 In the talk

• It lowers the net benefit of insurance, as Arrow’s point does for first-party moral hazard: “a tension between risk spreading and risk creation”. It does not mean insurance is bad, only “not as good as we might have thought” (paraphrase of).
• Magnitude: many people take slightly less care, versus a few who deliberately create loss. “you have to be kind of sociopathic to start off with”, yet the second effect “could be of us of the same magnitude” (the captions garble this; paraphrase) as the first. And it “garnered no real attention at all”.
• Three cost types: primary losses, secondary losses (creating and policing the scheme), and adverse selection from higher premiums.
• It is “largely extra contractual”: “there’s no privity between the insurer and the loss causer”.
• The consequence: “insurance is just not pure contract”. You may need “the police department or something else” on top, and “government as governance”.
• Loss control may merely move the loss: the kidnap wall “just shifts the loss down the block to the guy who doesn’t have a big wall in front of his house”.

📚 Deeper theory

1. The welfare accounting is idea 1: net benefit = risk-spreading gain minus the cost of changed behaviour. The new item is the behaviour of outsiders.
2. Asymmetry of numbers. Many small first-party changes versus few large third-party predations: the comparison is about the tails. A small number of well-organised schemes (the New York ring) can generate large claims, whereas a diffuse reduction in care is spread thin.
3. Why contract tools fail is Roots: all five insurer levers (underwriting, conditions, cost-sharing, claims handling, exclusions) act on the policyholder or the claims process.
4. Displacement is idea 6. A policyholder who builds a wall gains privately; the social loss is unchanged. In the first-party case, loss control that lowers a hurricane’s damage is a clear gain.

📚 Roots: Insurance as governance, and law’s old defences against third-party manipulation

The problem. Insurers are not only risk-bearers; they also shape what insured people do. How far can private contracts do the work of regulators, and where do they need the state?

The point. The paper’s main institutional claim is the one in the sentence “insurance is just not pure contract”. Insurance-as-governance literature says the contract is a governance tool. Siegelman says the tool is aimed at the wrong person when the manipulator is outside the contract. So the toolkit must include governance by the state (“government as governance”) or by market design (Lloyd’s). The second claim is about how governance can fail even when it works: the insurer who requires a higher wall does not reduce kidnapping, it relocates it. This is “private regulation with externalities”.

Thinker chain.

• Richard Ericson, Aaron Doyle and Dean Barry (2003), Insurance as Governance, is the standard sociological statement: insurers sort, price, monitor and reform conduct. Siegelman gestures at this literature (“there’s all kinds of takes on insurance as governance”).
• Omri Ben-Shahar and Kyle Logue (2012), “Outsourcing Regulation: How Insurance Reduces Moral Hazard”, argued that insurers can be better regulators than the state for some risks. Siegelman’s paper argues the converse for third-party moral hazard: insurers are weakest exactly where the loss causer is outside the contract.
• Baker and Griffith (2010), Ensuring Corporate Misconduct, on D&O insurance, which Siegelman uses for the D&O evidence.
• Anja Shortland (2019), Kidnap, on Lloyd’s special-risks underwriters acting as a private governance system for ransom markets. Baker and Shortland have a paper on kidnap and cyber ransom.
• The older legal doctrines: insurable interest (in England the Life Assurance Act 1774 and gambling rules; in the US, for example, Grigsby v. Russell, 1911) is a defence against third parties profiting from other people’s losses: you may not take out cover on a stranger’s life. Subrogation lets the insurer sue the wrongdoer. Anti-kickback and self-referral laws in health care (US federal law from the 1970s to 1990s) restrict providers’ financial ties. Each is a rule built because someone saw an outsider benefit from an insured loss.

Core argument in more detail. Ask what the insurer controls. It controls (1) who is insured (underwriting), (2) what the insured must do (conditions, warranties), (3) how much the insured bears (deductibles), (4) how claims are handled (adjusting, defence control), (5) what it will pay for (exclusions, sub-limits). All five are levers on the policyholder or on the claims process. A third party sits outside all five. The claimant in a staged crash is a claimant against someone else’s policy. The repair shop is a vendor chosen by the policyholder. The rehab centre bills the policyholder’s plan. The insurer’s only direct leverage is over the claims process (denial) and over payments to vendors (networks, pre-authorisation, “preferred provider” arrangements). Baker names those: managed care and preferred repair networks “are examples of” dealing with third-party moral hazard.

The old legal defences show a pattern. Each was a prohibition or status rule (insurable interest; no kickbacks; policy limits) rather than a contract term, because the manipulators are outside the contract. This fits Maher’s historical point that subrogation began in equity, not in contract.

Where others push back

• Insurers have more reach than Siegelman suggests: special investigation units, provider networks, claims data pooled across firms. Baker notes that managed care and preferred repair networks are, in effect, responses.

What they leave out

• Who pays for the “bigger” thing. Police resources for staged accidents are public money spent to protect a private pool; the funding question is not examined.

Check yourself

1. Why would a higher deductible not stop the staged-crash gang?
2. In what sense can loss control be “social waste” in the kidnap case?
3. Name the three categories of cost he lists.

---

3. Under-litigation, EPLI and the good-claim problem

In short. Siegelman says standard models assume frictionless litigation, but there is a massive under-litigation problem. In a hospital study he cites, about 17 per cent of patients had a serious adverse event and about 7 per cent of those injured claimed. So more claims driven by insurance could be a good thing. Insurers push anti-harassment training because it bulletproofs the policyholder against litigation, which shifts the cost back to the victim.

🎙 In the talk

• EPLI: anti-harassment training is often ineffective, but insurers push it anyway because “it bulletproofs the policyholder against litigation”. The cost of harassment is shifted back to the victim, who “can’t succeed in litigating”.
• Standard models assume frictionless litigation and error-free courts; in fact there is “a massive under litigation problem”.
• The hospital study he cites: about 17 per cent of patients had a serious adverse event, about 7 per cent of those injured claimed (as the captions give the figures).
• So more claims from insurance “could very well be a good thing” (paraphrase of). Both kinds of error exist: bad suits brought and good suits not.

📚 Deeper theory

1. The claiming pyramid (Felstiner, Abel and Sarat; Roots): most injuries never become claims.
2. Compliance as shield. Talesh’s work (cited by Siegelman; details from memory)† argues that insurers shape what counts as legal compliance. If an insurer-approved training programme counts as proof of diligence, the training becomes evidence for a defence, not a harm-reduction device.
3. Selection of disputes (Priest and Klein): what you see in court is not a random sample, so claim counts are a poor guide to the state of the world.
4. Type-one and type-two errors in claims handling. The same insurer incentive that denies bad claims also denies good ones.

📚 Roots: Tort, deterrence and the economics of claiming

The problem. Do injury victims actually claim? When they do, do claims track real harm? How should one judge a rise in claims: as fraud or as long-overdue access to compensation?

The point. The paper’s careful caution that more claims is not automatically bad draws on this tradition. The standard moral-hazard model, Siegelman says, assumes “litigation is frictionless and costless” and that “courts are error free”. The empirical picture is the opposite. In a world of “a massive under litigation problem”, more claims from insurance might help: insurance gives a victim a solvent defendant and a lawyer a reason to take the case. The honest position is that both bad suits are brought and good suits are not.

Thinker chain.

• Guido Calabresi (1970), The Costs of Accidents: tort has to reduce the sum of accident costs, avoidance costs and administration costs; the cheapest-cost-avoider idea. Steven Shavell (1980, 1987) and Landes and Posner (1987) built the economic theory of tort deterrence.
• Marc Galanter (1974), “Why the Haves Come Out Ahead”: repeat players do better. Felstiner, Abel and Sarat (1980–81), “Naming, Blaming, Claiming”: disputes are a pyramid; most injuries never become claims.
• The Harvard Medical Practice Study (Brennan, Leape, Localio and colleagues, early 1990s): many more patients were injured by negligent care than ever filed a claim. Siegelman’s figure (a study of a large academic hospital) belongs to the same family of work: “17 of patients” had a serious adverse event, “only seven percent” of the injured claimed. (Figures as the captions give them; the study is, I believe, by Lori Andrews and colleagues; not checked.)
• George Priest and Benjamin Klein (1984), “The Selection of Disputes for Litigation”: cases that reach court are not a random sample; both sides’ choices select them. Lucian Bebchuk (1996) on negative-expected-value suits: some suits are worth filing only because defence costs exceed the settlement.
• The False Claims Act (1863, amended 1986) is the US template for private enforcers who take a share of recoveries. Medicaid and Medicare fraud units use similar tools. Siegelman points to Medicaid for the qui tam proposal.

Core argument in more detail. Suppose the set of true harms is H. A fraction of them becomes claims; some claims are false. Insurance can raise both the true-claim rate (by making recovery worth the lawyer’s time) and the false-claim rate (by making the false claim worth the effort). A rise in total claims tells you nothing about welfare. A deterrence view says the key question is whether the claims go to the right defendants at the right level: do they push tortfeasors to take more care? Here Siegelman’s new point is that insurance on the tortfeasor’s side may weaken deterrence, because the insured tortfeasor pays only through premiums. The same worry appears in reverse in Maher’s ERISA point. If the victim’s own health plan takes the first dollar of any tort recovery, the victim and lawyer lose the incentive to sue at all, and the tortfeasor escapes. In his reading, “third-party moral hazard” also includes a more careless tortfeasor who expects that the victim’s insurer will pick up the bill and that nobody will sue. That is a deterrence story, not a fraud story.

Where others push back

• The malpractice figures are old and medicine-specific. In other lines (auto injury in some jurisdictions) over-claiming is the commoner complaint.

What they leave out

• Any way to measure the true baseline of harm outside medicine.

Check yourself

1. Why does the under-litigation point complicate any attempt to cut claims?
2. How does EPLI training shift costs back to victims, on his account?
3. Can fraud and under-litigation coexist? What does he say?

---

4. Kidnap and cyber ransom

In short. Siegelman says more kidnap insurance means more money, higher ransoms and a rising supply of kidnapping. Cyber ransom works the same way, as insurance makes firms more attractive targets. Kidnaps fell in Italy and Colombia after bans on insurance, but he is not 100 per cent confident since Italy did many things at once. A soft cartel at Lloyd’s limits cover so ransoms are paid out of pocket.

🎙 In the talk

• The theory: more insurance means more money, higher ransoms, and a rising supply curve of kidnapping. Kidnappers “pay attention to ransom size and the presence of insurance” (paraphrase of).
• Cyber is similar: more insurance makes people “more attractive targets” and encourages entry.
• Italy and Colombia: kidnaps declined after a ban on kidnap insurance, but “italy did a whole bunch of things at once”, and the Red Brigades may matter more. He says he is “not 100 confident”.
• Shortland’s account: a “soft cartel of insurance providers” at Lloyd’s who monitor each other, stop overpayment and limit coverage so that ransoms “have to be provided out of pocket”. They cover negotiation and business-interruption costs.
• Not covering the ransom, only clean-up costs, “actually is a potential benefit rather than a cost”.

📚 Deeper theory

1. Time-inconsistency of no-concessions (Schelling; Lapan and Sandler; Roots): insurance makes paying easy and so raises the incentive to take hostages.
2. Buyer concentration. A cartel of buyers acts like a monopsony: it can hold prices down. This is the reverse of the usual cartel worry.
3. Why exclude the ransom? Excluding the payment restores skin in the game for the victim and keeps the insurer from being the price-setter. It also fits idea 1: coinsurance on the thing most subject to manipulation.
4. Cyber differs. More firms are covered, the market is not concentrated at one floor, and many attacks are by actors outside any state’s reach. The Part 1 debate shows regulators and insurers split. The classic argument against a ban is that it hurts small firms and pushes attackers to larger targets (search summary; source not confirmed).

📚 Roots: Ransom, extortion and the credibility of non-payment

The problem. If a payer can be forced to pay, will the payment bring more demands? A government or firm would like to commit never to pay; but when the victim is already taken, paying is attractive. How do you make “no concessions” credible?

The point. Kidnap and cyber ransom are the paper’s clearest cases of an insured payment that is the product: the thing being sold by the extortionist is the very thing the policy covers. Siegelman gives the supply-curve argument: “when you have insurance you have more money available and higher ransoms”, higher ransoms mean more supply of kidnapping. He then gives Shortland’s account of a market that seems to contain the problem: a “soft cartel” of underwriters on the same floor of Lloyd’s, monitoring each other, refusing to overpay, and not covering the ransom itself.

Thinker chain.

• Thomas Schelling (1960), The Strategy of Conflict: commitments, threats and why tying your own hands can help. A credible promise not to pay removes the reason to take hostages.
• Harvey Lapan and Todd Sandler (1988), “To Bargain or Not to Bargain: That Is the Question” (American Economic Review, papers issue): the time-inconsistency of a no-concessions policy.
• Public policy: “no concessions” doctrines, and laws that restrict paying or freezing assets, such as Italy’s 1991 measure on victims’ assets (†, and I believe it froze assets rather than banned insurance).
• Anja Shortland (2019), Kidnap: the Lloyd’s special-risks market as an organised, concentrated buyer that polices prices and process (OUP book page exists; the description of its argument is from a search summary, source not confirmed).
• Ransomware economics: since the late 2010s, a literature and policy debate on whether insurers who cover extortion payments fuel the market (see Part 1 current debate and the TechTarget piece).

Core argument in more detail. Schelling’s point is that the victim’s best policy before the event (never pay) differs from the best policy after (pay to save the hostage). Insurance deepens this time-inconsistency in two ways. It makes the victim able to pay, which raises the price a kidnapper can demand. And it makes the payer an institution with no emotional stake. The insurer’s own interest is in a lower ransom; its power is that it is a repeat player that can coordinate. A single buyer in a one-off kidnap is a price-taker. A cartel of buyers that shares information and sets norms is a price-maker. That is the sense in which a “soft cartel” can reduce, not increase, third-party moral hazard, and why Siegelman reads Shortland as showing a design that “works tolerably well” (his phrase paraphrased).

Where others push back

• Attributing falls in kidnapping to insurance bans is hard: Siegelman says so himself.

What they leave out

• How cover is priced for a repeat target.

Check yourself

1. Why might a market of insurers who exclude the ransom itself reduce third-party moral hazard?
2. Why is the Italian evidence weak?
3. Why might cyber be harder to govern than kidnap?

---

5. Health care: the wallet biopsy, MRIs and copay coupons

In short. Siegelman rejects the claim that health care is expensive because patients overuse it, calling it ridiculously implausible. The real actors are hospitals, drug companies and doctors, who siphon off money flowing through the pipeline. His examples are a wallet biopsy, MRIs for uncomplicated back pain, and copay coupons that add several billion dollars in spending. Doctors are no better or worse than anybody else.

🎙 In the talk

• Hyman and Silver: “health care is expensive because it is insured”. Siegelman says that as a story of patient overuse it seems “ridiculously implausible” to him, and cites Nyman for the view that that if people get transplants only because of insurance it is “a good kind of moral hazard”.
• The real actors are “the hospitals it’s the drug companies it’s the doctors”. They “siphon off some of the money that’s flowing through the pipeline”.
• “a wallet biopsy”: treatment sized to the patient’s coverage.
• MRIs for uncomplicated back pain are ordered “without any clear clinical indication”; doctors may own the imaging facility.
• Arrow’s hope that doctors would “rely on the professional ethics of physicians”. Siegelman’s reply is that doctors are “no better or worse than anybody else”.
• Copay coupons: brand makers advertise the drug “available for as little as zero dollars”. The effect is “several billions of dollars” in extra spending; Massachusetts bans them.

📚 Deeper theory

1. Supplier-induced demand (Evans, Fuchs; Roots). The classic question: who decides how much care is bought? Insurance removes the price check on the supplier’s advice.
2. Conflict-of-interest law. The Stark and Anti-Kickback regimes exist because the problem Siegelman describes was seen decades ago†.
3. How a coupon defeats a copay. Tiered cost-sharing is the first-party fix. A coupon is a third-party counter-move that exploits it (Roots).
4. The insurer’s counter-counter-move. “Accumulator” programmes (search summary; source not confirmed) refuse to count coupon money toward the patient’s deductible. That is an arms race, not a solution.

📚 Roots: Health-care economics: supplier-induced demand and the “price capture” side of insurance

The problem. Health care is unusual because the doctor advises the patient and also sells to the patient, and a third party pays. Who decides how much care is enough?

The point. This tradition supplies the “price capture” sub-type. When Siegelman says “it’s the hospitals it’s the drug companies it’s the doctors”, he reframes the conventional story that patients overuse care. The overuse, he says, is mostly induced by suppliers who “siphon off some of the money that’s flowing through the pipeline”. The wallet-biopsy practice, MRIs “without any clear clinical indication” (as he reads from the reports), and copay coupons are the examples.

Thinker chain.

• Arrow (1963) again: trust and professional ethics stand in for markets. He expected doctors not to “cheat even though it may be rational economic behavior” (Siegelman reads this passage out).
• Robert Evans (1974) and Victor Fuchs (1978) on supplier-induced demand: providers can shift demand when they have information the patient lacks. John Wennberg and Alan Gittelsohn (1973) showed large, unexplained regional differences in practice style.
• Regulation of conflicts of interest: the federal Anti-Kickback Statute (1972) and the Stark self-referral laws (1989, 1993), aimed at physicians who profit from referring patients to facilities they own. Siegelman’s remark that “doctors sometimes own the imaging facility” is exactly the Stark pattern.
• Pharmaceutical pricing: the Hatch-Waxman Act (1984) created the modern generic pathway; tiered formularies and pharmacy-benefit managers steer patients to generics with copay differences. Research on coupons: Ross and Kesselheim (2013) in the New England Journal of Medicine argued that coupons can raise spending by undermining formulary incentives, and later economic studies measured effects on generic use†. Coupons are not allowed for people on federal programmes, because of the Anti-Kickback Statute†.

Core argument in more detail. The coupon case is the cleanest. The insurer designs a price signal: low copay for generic, high for brand. The patient prefers the brand. The brand maker offers a coupon that cancels the difference, so the patient pays the same either way, and the insurer’s signal is turned off. The insurer is now paying the markup on the brand. The coupon works only because the insurer’s cost-sharing exists. Remove cost-sharing and the coupon has nothing to cancel. So the instrument that the first-party theory recommends (cost-sharing) gives the third party a handle. This is why the usual first-party tools can backfire in the third-party setting (my gloss, from a search summary of the abstract, not from the talk).

Arrow’s sentence matters for a different reason: it shows that the founding paper on moral hazard itself relied on an informal institution (professional ethics) to guard against third-party-type behaviour, and that the institution was assumed, not studied. Siegelman’s retort is deliberately gentle: if doctors were what Arrow wanted them to be, “the world would be a better place”.

The sub-type also explains why the term “moral hazard” can mislead. The patient who gets an MRI because it is free is a first-party effect; the doctor who orders it because he owns the scanner is a third-party effect. Nyman’s argument that some first-party response is good (access) makes it important to pull the two apart, as Siegelman does.

Where others push back

• Defenders say coupons help patients afford drugs; the host raises the idea that coupons could be defended as a way of funding innovation.

What they leave out

• Single-payer or tightly managed systems, where the third parties still exist but the pool and the rules differ. For India, a state-funded scheme that pays hospitals per package faces the “price capture” and “volume” version of this problem; I believe schemes of that kind have anti-fraud units, but I did not verify this†.

Check yourself

1. Why does Siegelman think the idea that patients overuse because care is free is the wrong main story?
2. How does a coupon undo a tiered copay?
3. What is the irony in Arrow’s remedy?

---

6. Liability insurance: no-fault billing rings and D&O

In short. Siegelman again says the paper does not allege a mass outbreak of fraudulent lawsuits. His main case is a New York indictment of organised-crime figures who padded bills under the state’s no-fault law, with runners finding accident victims and their own clinics billing at every stage. On D&O insurance, one interviewee said it is easier to get money from an insurer because it’s the third party’s money, but he calls that evidence weaker.

🎙 In the talk

• Again the baseline caveat: there is “under claiming” and the paper does not allege “a mass outbreak of fraudulent lawsuits” (paraphrase of).
• A federal indictment in New York: Russian organised-crime figures padding bills under the state’s no-fault law. Runners found accident victims and got “a commission for each person”; their own ambulances, clinics, MRI shops and physical therapists billed at every stage.
• D&O: Baker and Griffith’s qualitative study; one interviewee said it is easier to get money from an insurer “because it’s the third party’s money”. Siegelman says the evidence is “weaker”.

📚 Deeper theory

1. Vertical integration of a fraud chain. Each node refers to the next and bills at the maximum: Roots on organised, staged crime; the criminology of insurance fraud rings.
2. No-fault design. No-fault systems pay medical bills quickly regardless of fault. The design removes the fault-finding check; it also means the payer is exposed to billing at every step†.
3. D&O and strike suits. Baker and Griffith (2010) found that D&O is a place where insurers rarely police defence costs and where the payment of settlements is largely by insurers†.

📚 Roots: Crime, opportunity and displacement

The problem. How do you reduce crime: change offenders, or change opportunities? And does blocking one route simply send offenders elsewhere?

The point. The staged crash and the New York no-fault ring are crime stories; the insurance policy is the “suitable target” with no “guardian”. Siegelman’s kidnap wall is displacement in its simplest form: “it just shifts the loss down the block to the guy who doesn’t have a big wall in front of his house”. And the remark that “you have to be kind of sociopathic to start off with” is the motivated-offender condition: only a few act, yet a few suffice if the target is rich.

Thinker chain.

• Gary Becker (1968), “Crime and Punishment”: offenders weigh expected gain against expected penalty. Cost-benefit crime.
• Lawrence Cohen and Marcus Felson (1979), routine activity theory: crime needs a motivated offender, a suitable target and the absence of a capable guardian, meeting in time and space. Change in everyday life changes opportunity.
• Ronald Clarke and colleagues (1980s onward), situational crime prevention: make crime harder, riskier, less rewarding. Reppetto (1976) and later reviews on displacement: blocking one target may shift crime to another place or method, though evidence shows displacement is often partial, not total†.
• Insurance-fraud research: Richard Derrig and others, on “soft” (opportunistic padding) and “hard” (planned) fraud in auto claims, and on organised rings†. Tennyson on public attitudes to insurance fraud†.
• Behavioural ethics: Mazar, Amir and Ariely (2008) found that people cheat a little, but not as much as they could, to keep a self-image of honesty. Impersonal victims make a little cheating easier.

Core argument in more detail. Routine-activity theory gives a neat map. Insurance changes the target (a deep pocket, with identity-free cover) and removes a guardian (the person who would have defended his own money). The offender’s gain is large and the victim is impersonal, so the moral cost to the offender drops, as Baker’s “blood money” evidence suggests: lawyers feel differently about taking money from “mrs jones” than from a company. The result is a rise in opportunity crime that has a distinct signature: it is organised (runners, ambulances, clinics, MRI shops, physical therapists, as in the New York ring), staged, and often legal-looking.

The criminology also helps with the “something bigger” answer. If the guardian is missing, you can supply one: police, cameras, data sharing between insurers (the Lloyd’s practice of insurers “monitoring each other”), or paid private guardians such as bounty hunters. Each is a form of guardianship by someone other than the insurer.

Where others push back

• An indictment is one case. The size of the problem across no-fault systems is unclear.

What they leave out

• Regulatory responses within no-fault states, such as fee schedules, and what they did to the rings.

Check yourself

1. What made the New York scheme “organised” in the literal sense?
2. Why is the D&O evidence weaker than the no-fault evidence?
3. How does no-fault design change who checks a bill?

---

7. Mechanisms: deep pockets, impersonal payers, crowded-out monitors

In short. Siegelman gives mechanisms. The simplest is deep pockets. Plaintiff lawyers almost never pursue more than the policy limits, because taking from an insurer feels different from taking from Mrs Jones. Insurance can also crowd out monitoring: an insured car owner visits fewer repair shops. In finance, government-backed insurance eroded the discipline lenders would otherwise impose on banks.

🎙 In the talk

• The simplest story: “deep pockets”.
• Baker’s “blood money” paper: plaintiff-side lawyers almost never go after more than the policy limits, because “that just feels different” to take money from “mrs jones” than from an insurer.
• Insurance can crowd out monitoring: “insurance crowding out monitoring”. A car-repair customer who is paying would visit five shops; an insured one visits fewer.
• Finance: private creditors of too-big-to-fail banks did not watch the banks’ loans closely. “the implicit promise of government-backed insurance eroded market-based incentives for financial discipline by lenders”.

📚 Deeper theory

1. Rents and contests (idea 1): a bigger prize draws more effort.
2. Moral psychology. People cheat more when the victim is abstract (Mazar, Amir and Ariely; idea 6)†. Baker’s “blood money” is evidence from legal practice.
3. Monitor chains (Roots): guarantees relax the whole chain of watchers.
4. Policy limits as a focal point. The norm of not going beyond the policy limit works like a price ceiling on the plaintiff’s demand. It may reduce harsh outcomes for individual defendants while also shaping what claims are brought.

📚 Roots: Safety nets, bailouts and the erosion of market monitoring

The problem. If the state or an insurer stands behind a firm, does the firm’s own creditors’ watchfulness fade? That is, does protection reduce other people’s discipline of the protected party?

The point. The mechanism Siegelman takes from finance is the third of his three: insurance can crowd out monitoring by others. The too-big-to-fail creditors argued, in his words, that “the implicit promise of government-backed insurance eroded market-based incentives for financial discipline by lenders”. The first-party story is that the bank takes risks. The third-party story is that its creditors stop watching. The car-repair example has the same structure in miniature: the insured customer stops shopping around, so the repair shop faces less price discipline.

Thinker chain.

• Diamond and Dybvig (1983) on bank runs and deposit insurance as a fix, with the known price: moral hazard. Kareken and Wallace (1978) made the case that deposit insurance without risk-based pricing invites risk-taking.
• Edward Kane (1985), The Gathering Crisis in Federal Deposit Insurance: the 1980s savings-and-loan collapse as an incentive failure.
• The “too big to fail” literature (for example Stern and Feldman, 2004) and studies of market discipline on banks, including work after 2008 asking whether creditors price in the implicit guarantee (for example Acharya, Anginer and Warburton on lower funding costs for large banks)†.

Core argument in more detail. Think of a firm’s behaviour as constrained by a chain of monitors: shareholders, creditors, customers, regulators. Each monitor exists because each is exposed to the firm’s failure. A guarantee that covers one link (deposit insurance, a repair bill) relaxes the other links too, even though they are not themselves insured in any formal sense, because their exposure falls. So a guarantee does two things: it lowers the protected party’s own care (first-party effect) and it lowers the watchfulness of everyone whose money rides on the protected party (third-party effect). The second effect is invisible in a model with a single decision-maker, and that is exactly why it had no name.

Where others push back

• The finance evidence is contested and bank guarantees are not insurance contracts.

What they leave out

• Whether the feels-different effect is changing as claims handling gets more automated.

Check yourself

1. Why would a lawyer not pursue assets above a policy limit?
2. What is “monitoring” and how can insurance crowd it out?
3. Why is the finance case a third-party, not a first-party, story?

---

8. Remedies: market design, the duty to reimburse, bounties

In short. Siegelman admits he has no great ideas but lists remedies that exist. These include the Lloyd’s kidnap market, the D&O duty to reimburse as a pre-commitment device, subsidies for effective precautions such as dashboard cameras, and qui tam bounties paid to those who recover from fraudsters. His summary is to look at insurance in a general equilibrium context.

🎙 In the talk

• “we don’t have any great ideas”. He lists solutions that exist.
• Lloyd’s kidnap market (idea 4).
• The D&O duty to reimburse as “a powerful sort of pre-commitment device”: a plaintiff who sees a defendant defending vigorously with someone else’s money may not bring a strike suit. The puzzle it answers is why D&O policies are so generous on defence terms. He says he has no “direct evidence” of that.
• Subsidise effective precautions (dashboard cameras in Europe; not in the US as far as he knows).
• Qui tam bounties: insurers pay private actors a share of what they recover from fraudsters, modelled on Medicaid.
• The summary: look at insurance “in a general equilibrium context”.

📚 Deeper theory

1. Commitment devices (Schelling; idea 4): precommitting to a vigorous defence changes the plaintiff’s calculation.
2. Why bounties. A private enforcer fills the missing guardian (idea 6) and avoids giving the insurer more power to deny (idea 2). The False Claims Act experience gives a template and a warning: bounty-driven suits can themselves produce over-enforcement†.
3. Subsidised precautions are Pigovian in the other direction: pay for the loss-reducing activity.
4. Combining remedies. Market design (concentrate buyers), conduct rules (kickback bans) and enforcement (bounties) address different failure points.

Where others push back

• A duty to reimburse also invites first-party abuse (running up defence costs), a point he concedes.

What they leave out

• Costs of administering a bounty programme and the risk of false accusations.

Check yourself

1. How can a generous defence policy deter a lawsuit?
2. What problem do qui tam bounties solve that denial rules cannot?
3. Why does he end on “general equilibrium”?

---

9. The Q&A: subrogation, gradations, ideology and ERISA

In short. In the Q&A, Maher says subrogation began as an equitable right and Siegelman calls it a tool useful only within limits. He calls his work anecdote mongering and admits a biased sample. Baker urges subcategories, and Siegelman offers the milder kinds as the might-as-well and can’t-be-bothered moral hazard. He warns the term has an ideological valence, and says he can say what makes something inefficient but has no grounding in moral philosophy.

🎙 In the talk

• Maher on subrogation: it began as an equitable right, so “150 200 years ago common law judges had the exact same intuition that you have now”. Siegelman: it is part of the toolkit, “useful only within limits”, because the actors are “judgment-proof” or hard to find.
• Fraud versus negligence: he calls his work “anecdote mongering” and admits “i have a biased sample”; he does not know the split between bad actors and extra negligence.
• Baker: once you name something, people see it. He asks for subcategories: a doctor raising prices feels different from the Florida shuffle. He calls the milder kind “the might as well moral hazard” and “can’t be bothered moral hazard”. Naming might legitimise managed care. He warns that “moral hazard has an ideological valence”. Siegelman says he has no position on whether insurers are “getting ripped off” (paraphrase of).
• Host: how to separate good from bad; intent; copay coupons. Siegelman: “i have no grounding in moral philosophy”, and “i can tell you what makes something inefficient”.
• Yass: Florida assignment of benefits for roofing and auto glass, where attorney fees drive cost. Siegelman had not heard of it.
• Maher on ERISA: first-dollar reimbursement clauses that are “not even subject to the common fund exception” leave victims and lawyers with no reason to sue the tortfeasor. Siegelman: “can barely understand” it but is glad the tool is being used.

📚 Deeper theory

1. Subrogation as a legal tool: a right of the insurer against the wrongdoer. It works where the wrongdoer is identifiable and solvent (negligent drivers) and fails where the actor is hidden, judgment-proof or looks legitimate (idea 2).
2. Gradations. Baker’s sub-categories map onto the three sub-types I gave in Part 1 (predation, price capture, crowded-out monitoring), and onto a moral gradient. A legal system can respond differently: criminal law for predation, consumer or antitrust law for price capture, contract design for monitoring.
3. Ideology of the term. “Moral hazard” in political argument tends to justify tighter contract terms, as Baker says. A concept that shifts blame to outsiders could be used the same way. Naming has costs.
4. Assignment of benefits is a classic third-party pattern: the contractor takes the policyholder’s claim right and the dispute becomes a fee-driven litigation. (Details of Florida’s rules are†)

Where others push back

• On the host’s good-versus-bad question, Siegelman’s answer is that the paper deliberately avoids drawing the line. Critics will say a concept without a normative test cannot guide law.

What they leave out

• What a test would look like. Baker’s suggestion to use legal and historical judgments of “bad” is left as future work.

Check yourself

1. Why is subrogation weak against the examples in the paper?
2. What is the difference between “might as well” and “Florida shuffle” cases, and why does it matter for policy?
3. How does the ERISA point show third-party moral hazard working through the victim’s incentives?

---

About the source and how this page was made

• Speaker: Peter Siegelman (with co-author Gideon Parchomovsky, not on the call)
• Event: UConn Law School Insurance Law Center workshop series, online (Webex); Q&A with Brendan Maher, Tom Baker and others
• Date: 2022-04-07 (from the video description; uploaded 2022-04-25)
• Length: 1:06:56

Verification note. I checked a few links by fetching the page and confirming that it contains the words I rely on: the Journal of Legal Studies page for the paper, the Wikipedia entry for the Florida shuffle (it lists the New Yorker piece), the Littler note on Montanile, the TechTarget piece on cyber ransom cover, the SCC Online headline on fake motor claims, and the EconPapers entry for “The Dark Side of Insurance”. Anything marked “(search summary; source not confirmed)” came from a web-search summary that I could not tie to a page. Anything marked ”†” is my background knowledge, which may contain errors. Dates, case names and holdings in Part 2 are mostly of that second kind. The talk gives almost no hard numbers, and I have added none of my own.

Speakers

Label	Who	Role
Host	Not named in the captions. Others address the host as “Travis” (caption spelling may be wrong)	Runs the UConn insurance-law workshop; introduces the talk, keeps the queue, asks a late question
Siegelman	Peter Siegelman	Presenter. The host calls him the “philip blungberg professor of law” at UConn Law (caption spelling of Phillip Blumberg) and says “his own paper”. Co-author of the paper with Gideon Parchomovsky
Parchomovsky	Gideon Parchomovsky	Co-author, named in the video description. Not on the call
Maher	Brendan Maher	Asks about subrogation and about how much of the problem is fraud and how much is negligence; closes with an ERISA point. Camera not working, voice only
Baker	Tom Baker	Insurance-law scholar (cited by Siegelman as “tom baker”). Offers a comment on gradations of the problem and on the word “moral hazard”
Yass	”bob yass” (caption spelling)	Asks about Florida assignment-of-benefits claims
Host (late)	As above	Asks how to separate “good” from “bad” third-party moral hazard

Pronouns: only Siegelman is “he” (the host says “his own paper”). For the others I use surnames.

Caption warnings. Auto-captions. Names I corrected on this page, from memory and context: “philip blungberg” = Phillip Blumberg; “carolinkowski” = probably Carolyn Kousky (flood insurance)†; “kyle logg / kyle logan” = Kyle Logue; “shaheen teleth” = Shauhin Talesh; “anya shortland” = Anja Shortland; “ketam” = qui tam; “dno” = D&O (directors’ and officers’ liability); “epli” = EPLI (employment practices liability insurance); “ronin offram and ariel perez” = Ronen Avraham and Ariel Porat, “The Dark Side of Insurance” (the second author’s name is confirmed by search; the captions garble both names); “laurie andrews” = probably Lori Andrews†; “john nyman” is right; “daftboard cameras” = dashboard cameras; “tort visa / torpedoes” = tortfeasors; “segregation” in the Q&A = subrogation; “logan schneiderman” in the kidnap section is garbled and I could not identify it; “first dollary” = first-dollar. The 17% and 7% malpractice figures are as the captions give them.

Factual points to treat with care.

1. The paper is described as “forthcoming” in the Journal of Legal Studies. The journal’s page lists it in vol. 51, issue 1 (confirmed); the year and page range (2022, 93–131) are from a search summary, not confirmed. The talk date, 7 April 2022, comes from the video description. A scheduling gap between “forthcoming” and the issue date is not unusual; I do not treat it as an error.
2. At Baker half-remembers a film scene about HMOs and says the line was given to a doctor that Jack Nicholson paid for. From memory†, the famous HMO insult in As Good As It Gets is spoken by the Helen Hunt character, not by a doctor or by Nicholson. Baker himself is unsure (“it wasn’t jack nicklaus”). It does not affect his point.
3. At Siegelman says Italy and Colombia banned kidnap insurance and kidnaps fell. He hedges it himself, and so should we. From memory†, Italy’s 1991 measure froze the assets of kidnap victims’ families, which is not quite a ban on insurance, and Italy was also fighting political violence at the time.

---

🎙 marks what the speaker says. 📚 marks added background and the wider debate, not the speaker’s claim. † marks background from general knowledge that was not checked against a source.