Liability Too Late

Ex-post liability locks oil and gas governance into Hollow Win outcomes

The paper argues that liability after damage cannot preserve system welfare because the regulatory game is built around firm profit and host-state revenue. The ecological and climate system is absent at the extraction decision point.

• 1 predicts Hollow Win (0,1,1): C is degraded while A and B gain
• The Missing System Theorem explains why C is outside the bilateral game G = A, B
• Liability payments occur after the decision D_t and after system damage has begun

The payoff space contains BP and the state, not the Gulf system

A1 defines the oil and gas regulatory game as a bilateral payoff space: firm profit π and state revenue π . Deepwater Horizon shows how a cap can be normal inside that game while absurd relative to system loss.

• BP held a 65% stake in Macondo, with Transocean as rig owner and Halliburton as cementing contractor
• BP's Gulf of Mexico operations generated about $12 billion in annual regional revenue
• The federal government collected $5.4 billion in offshore oil and gas royalties in fiscal year 2009
• OPA's $75 million offshore facility damages cap bounded the bilateral decision space

Liability arrives at t2 after the system has already been damaged

A2 makes timing the central defect. Extraction is approved at t0, damage occurs at t1, and compensation is paid at t2. The system cannot wait for adjudication.

• CO2 remains in the atmosphere for 300 to 1,000 years
• Methane has 80 times the 20-year global warming potential of CO2
• Macondo discharged an estimated 4.9 million barrels over 87 days before the well was capped
• The paper calls this the Irreversibility Floor for climate channels

A finite cap changes the firm's cost function, not the damage

A3 says the firm optimizes against K when liability is capped. The expected cost becomes p · K, even when true system destruction is far larger.

• For Macondo, the statutory cap was $75 million
• The paper estimates Macondo system welfare destruction at about $81.7 billion
• The proof sketch also uses W_dest ≈ $60 billion to show K < W_dest
• Congress raised the cap to $134 million in 2016, leaving the same capped structure in place

The $81.7 billion Macondo estimate comes from six damage channels

The paper does not treat Deepwater Horizon as one cleanup bill. It adds spill response, natural resources, local economic loss, ecosystem services, health exposure, and methane climate damage.

• Direct cleanup costs: $14 billion
• Natural resource damages: $8.8 billion
• Gulf fisheries and tourism losses: $17.2 billion
• Long-term ecosystem degradation: $34 billion over 30 years
• Human health impacts: 6.5 billion, plus 1.2 billion from methane climate impact

Nigeria fits the paper's φ > 0.5 capture condition

A4 predicts capture when a host state gets more than half of its fiscal revenue from extraction. In the Niger Delta case, the state and Shell had aligned incentives to keep extraction moving and under-remediate spills.

• Nigeria relied on oil revenue for about 70% of its budget
• SPDC's Nigerian operations generated about $8.6 billion in annual revenue from 2000 to 2010
• The Nigerian federal government collected about $3.2 billion annually from taxes, royalties, and its 55% equity share
• NOSDRA is described as chronically underfunded and captured by joint-venture revenue interests

A low β_W can still hide concentrated regional destruction

The Niger Delta case has β_W = 0.043 because revenue is measured across decades and beneficiaries, while damage is concentrated in one region. The paper uses this to show why aggregate ratios can understate weakest-link harm.

• Estimated spills from 1970 to 2020: 7,000 to 13,000
• Estimated oil released: 3 to 5 million barrels
• UNEP found soil and groundwater contamination above standards by factors of 10 to 1,000 in 69% of examined Ogoniland sites
• Estimated system welfare destruction: about $18.3 billion

Damage moves toward the lowest expected liability cost

A5 states that in a global market, extraction concentrates where expected liability E[L_j] is minimized. The paper uses this mechanism to explain why strong rules in one jurisdiction do not set the global equilibrium.

• The relevant governance variable is g, the jurisdiction's governance capacity
• The lowest-g jurisdiction sets the weakest liability link
• The Niger Delta concentrates damage while revenue flows to SPDC and the federal government
• The theorem predicts migration unless system pricing is applied at the decision point

Single-channel liability leaves most system welfare unpriced

A6 defines total welfare destruction as an additive sum across independent channels. Addressing spill liability alone does not price climate, methane, health, ecological, or governance damage.

• W_total = W_CO2 + W_CH4 + W_AP + W_spill + W_gov
• OPA-style liability mainly addresses spill-related damages
• Climate damages are valued in the paper using the EPA's 2023 SCC of $190 per tonne
• A liability regime can pay one channel while the other channels keep accumulating

The paper's global oil and gas β_W is 2.0

β_W = -dW/dΠ measures system welfare destruction per dollar of industry revenue. The paper treats it as the headline diagnostic for whether liability is missing the real cost of extraction.

• Global annual oil and gas revenue Π is approximately $5 trillion
• Annual system welfare destruction exceeds $10 trillion
• The calculation includes climate damages, methane operational damages, air pollution health burdens, and ecological destruction
• β_W = 2.0 means 1 of revenue corresponds to 2 of system welfare destruction

System Welfare Bonds move accounting to the lease decision

Rule R changes G = A, B into G1 = A, B, C . Before a lease is granted, the operator posts collateral equal to the project's estimated β_W-adjusted welfare cost.

• The bond is posted before extraction, not after damage
• The bond is returned with interest only if the project achieves (1,1,1)
• The paper defines full success as carbon sequestration, zero spills, and local ecosystem preservation
• If the project degrades the system, the bond is forfeited to a system-welfare restoration fund

The theorem can be refuted, but only by cases the paper says have not been observed

The paper gives five falsification conditions. They are useful teaching anchors because they separate the theorem from a policy slogan.

• F1: an ex-post payment fully restores all climate, methane, health, ecology, and governance damage
• F2: expected liability equals W_dest for all scenarios with no cap, perfect information, and Δt = 0
• F3: a state with φ > 0.5 consistently preserves C despite extraction revenue dependence
• F4: extraction concentrates in high-governance jurisdictions despite lower-liability alternatives
• F5: a decision produces (1,1,1) without Decision Accounting Field 17