The Subsidy Race Has No Finish Line

Industrial policy returned to respectability quickly, and the return was justified. Some supply chains had concentrated to a degree that no serious government could accept, and the market had no mechanism for undoing that on its own.

What followed was less considered. Once several large economies began offering incentives for the same set of industries, the packages stopped being a response to a strategic gap and became a bidding process, with the firms as auctioneers.

The auction dynamic

A company evaluating locations does not have to lie or threaten. It only has to keep multiple jurisdictions in the process, which is ordinary corporate practice, and let each one learn what the others have offered.

The escalation is visible in the numbers. Incentive packages per job have climbed steadily, and the climb has not been matched by any improvement in the underlying business case, because the business case was never what was being bid on.

When several buyers compete for a good that only one seller can supply, the price is set by the second most eager buyer, not by the value of the good.

Clawbacks are the weak point

Nearly every package includes performance conditions. Jobs created, wages paid, investment sustained for a period of years. Enforcement of those conditions is where the system consistently fails.

The failure is political rather than legal. Enforcing a clawback means a public admission that the original deal was a mistake, usually by the successor of the official who signed it, against a company that still employs people locally. That is a fight nobody chooses.

What a disciplined version looks like

It exists, in a few places. Independent scoring of every package before signature, published. Automatic clawbacks that trigger without an official having to initiate them. Ceilings expressed per job rather than per project. Sunset dates on the programs themselves.

These do not end the competition. They make it a competition on terms that a government can afford to lose, which is the only version worth entering.

The case for the policy is real and should not be conceded away by its critics. A supply chain concentrated in one jurisdiction is a vulnerability no market corrects, because no individual firm bears the cost of the concentration.

What went wrong is not the objective but the auction that followed it. Once several large economies wanted the same plants, the packages stopped being a response to market failure and became bids.

The bidder's problem is structural. A firm choosing a location does not need to threaten anything. It only needs to keep several jurisdictions in the process, which is ordinary practice and entirely lawful.

The award data records what was granted, which makes the federal share of this unusually traceable compared with the state and local packages that often exceed it.

That asymmetry matters. The visible number is frequently the smaller one, and the incentives that decide a siting decision are often assembled from sources no single register captures.

The international layer is supposed to constrain this and does so weakly. The subsidies rules distinguish prohibited from actionable support and depend on a dispute mechanism that has to be functioning to bite.

Where that mechanism is degraded, the constraint becomes reputational, and reputation is not a strong brake on a government that has decided a plant is strategic.

The performance conditions are the part that would fix the worst of it, and they are the part that fails. Nearly every package includes them. Almost none are enforced.

The reason is political rather than legal. Triggering a clawback requires a public official to announce that a deal they or a predecessor signed did not work, in a place where the jobs were promised.

The budget documents carry the ongoing cost of these commitments, often as forgone revenue rather than as spending, which is the accounting treatment least likely to attract scrutiny.

The reforms that work are unglamorous and exist in a few places. Independent scoring before signature, published. Clawbacks that trigger automatically on a missed milestone rather than on a decision to pursue them.

The administering agency publishes enough to make comparison possible, and the comparison nobody makes is the one that matters: cost per job against the same figure five years later.