The asset with no outside

On Friday, 5 August 2011, Standard & Poor’s lowered the long-term credit rating of the United States from AAA to AA+. American sovereign debt had held the highest grade for roughly seventy years, and this was the first time it fell below it. The decision came days after Congress ended a debt-ceiling standoff with the Budget Control Act, and S&P rested its reasoning on the condition of American political institutions, which it described as having grown less stable, less effective, and less predictable, rather than on any new fact about the government’s capacity to pay.

The asset that had been downgraded rose in price. In the days after the announcement, yields on US Treasuries fell, which means investors paid more to hold the debt that had just been declared less safe. The European Central Bank, reviewing the episode in its Financial Stability Review of December 2011, recorded that the downgrade produced no clear reaction in the bond market: Treasury yields did not increase and market liquidity did not dry up. Money moved toward the instrument that had been marked down. The American stock market, which had not been downgraded, fell between five and seven percent in a single session.

The standard account of this is not wrong. August 2011 sat inside the European sovereign debt crisis, and capital was leaving the euro for the dollar. A downgrade of US debt, on this reading, was overwhelmed by a flight to safety that happened to terminate in the same US debt. The explanation is real, and it covers the direction of the flow. It does not cover the coherence of that flow with the downgrade. No quantity of flight from Europe explains why frightened money would run toward the very asset just declared riskier, unless the declaration carried no information about an external risk from which the money was fleeing.

A second fact complicates the picture in the other direction, and it should be stated rather than avoided. The cost of insuring US debt against default did rise. Credit default swap premia on Treasuries moved from roughly twenty-five basis points before the crisis to a range near fifty-five to seventy-five in 2011. A gauge of risk did register something. The phenomenon is therefore not that no price moved. The phenomenon is that the price of the downgraded asset moved in the direction opposite to what the downgrade implied, while a separate gauge moved in the implied direction, and the two were never reconciled. The market paid more to insure against American default and paid more to own American debt at the same time. Those two behaviours are incompatible if the rating measures a property of the asset. They are compatible only if the Treasury occupies a different position from the assets it is grouped with: the surface against which the riskiness of everything else is read.

The episode exposes a structural fact. A rating scale presupposes a riskless point. The grades are positions relative to it; to be rated is to be placed at some distance from a thing that is, by construction, the reference. In the dollar system, that reference is the US Treasury. It is the collateral that lubricates lending between institutions, the safe asset that portfolios hold to anchor everything riskier, the instrument treated as cash. To downgrade the Treasury is to apply the scale to the object the scale is built around. It resembles a ruler announcing that the standard metre is ninety centimetres long. The announcement does not record an error in the metre. It reveals that the scale loses its meaning when turned on its own foundation. The market grasped this in a weekend and did the only coherent thing available: facing uncertainty, it bought more of the asset just declared less safe, because there was no safer asset to move toward. The flight to safety and the downgrade pointed at the same instrument because the system has no outside.

The institutions that operate the rating confirmed the point within days, and in the most explicit way. Shortly after the downgrade, the Federal Reserve, the FDIC, and other bank regulators announced that for the purposes of bank capital and collateral, Treasuries would continue to be treated exactly as before. The rating had changed; the bodies that translate ratings into consequences declared that this particular change would not count. The creditworthiness of US debt, in the sense that governs the system’s behaviour, turned out to be a status the institutional order confers, and can withhold from the rating agency’s own verdict. S&P, for its part, had proceeded after the Treasury identified a two-trillion-dollar error in its debt projection. The verdict survived both the discovery of its arithmetic mistake and the refusal of the regulators to honour it.

A deeper question sits underneath, and the standard literature handles it poorly. If a rating can be arithmetically wrong, can be overridden by the institutions that use it, and in 2008 had stamped its highest grade on structured products that turned out to be worthless, why does the authority of the rating not collapse? The usual answers point to inertia, regulatory capture, the absence of alternatives, the cost of building another system. Each is true. Each explains why the rating is hard to displace. None explains why being demonstrably circular does not delegitimise it. Those are different questions. The first belongs to political economy. The second is about what kind of object a rating has become.

A rating presents itself as a statement about solvency, a claim that could in principle be checked against the solvency it describes and found accurate or mistaken. By 2011 it was no longer functioning that way. The rating had become a component of the solvency it claimed to measure: a downgrade raises borrowing costs, borrowing costs feed the deterioration, the deterioration ratifies the downgrade. Once a representation enters the object it represents, the question of accuracy loses its grip. Asking whether the rating is correct comes to resemble asking whether the market price of a thing is the right price. There is no independent value underneath the price against which to check it, because the price is what we mean by the value. The rating closed the same loop. It stopped pointing at an external referent and began to participate in producing the condition it reported.

A familiar account comes close to this and stops short of it. The performative reading of markets, developed by Callon and MacKenzie, holds that economic representations help to shape the objects they appear to describe: a pricing model, once adopted, bends the market toward the behaviour it assumes. That account keeps two terms in play, the representation and the object it acts upon, and traces a force running from the first to the second. The Treasury case removes the second term. There is no independent measure of the safety of the dollar system’s reference asset that the rating could be bending, because the reference asset is the thing safety is measured against. In this case the representation did not act on a referent standing outside it; the referent had been drawn inside, and nothing external remained against which the rating could be checked. That is a stronger condition than performativity describes, and it is the condition the 2011 episode made visible.

MacKenzie named the sharper face of the anomaly. He calls it counterperformativity: the use of a piece of economics makes the world conform less well to its own description. The downgrade fits the name, since S&P applied its scale and the market moved against what the scale implied. Counterperformativity still holds the relation at the level of influence; the model pushes the world one way or another and the world stays a separate thing being pushed. The Treasury case asks for the step the performativity literature has been careful to avoid, the claim that the device supplies its referent rather than pushing one that already exists. Critics of that literature have marked the same gap from the other side, observing that none of its grades describes a constitutive relation, the case in which a representation establishes part of the reality it names. That is the gap the closed device occupies.

This closure is the source of the authority, which is why the circularity does not corrode it. A measurement that answered to an external fact would be permanently contestable, revisable whenever the fact and the measurement diverged, and unstable as a basis for coordination. A device that has folded the referent inside itself offers something the honest measurement cannot: a fixed point. Millions of allocation decisions, capital rules, and mandates can anchor to it precisely because it no longer floats against an outside that might contradict it. The property that empties the rating of external truth is the property that makes it usable as shared infrastructure. It coordinates because it has stopped describing.

The structure is not peculiar to finance. It is the structure of the certificate that fixes the value of a work of art. The certificate of authenticity does not report a value the work already possesses and that an expert could verify independently; it constitutes that value, and because it constitutes it, the certificate cannot be wrong about an underlying worth, since no such worth exists apart from the act of certification. Provenance, attribution, and the institutional apparatus around them do not measure the art’s standing. They produce the standing they appear to record. The rating and the certificate are the same kind of device: an instrument that generates the reality it claims to observe, and whose authority rests on the closure itself. The circularity is the mechanism. It is what allows a single signature, or a single grade, to organise the behaviour of people who will never inspect the thing in question.

Once the device is seen clearly in these two cases, others fall into the same family. A price is such a device. A reputation is one. An index that compresses a country, a university, or a firm into a single rank is one. Each takes a field too large for any participant to survey, returns a single legible token, and then governs the field through the token it produced. The systems we are most tempted to read as descriptions, the ones that promise to tell us what something is worth or how safe it is, are frequently the ones that have closed the loop most completely, and so do the least describing and the most producing.

Two older bodies of thought stand behind all of this. Niklas Luhmann described modern society as a set of operationally closed systems, the economy among them, each producing its own elements through its own operations and reaching the rest of the world only as filtered information. Jean Baudrillard, writing about money once it had left the gold standard, described a sign cut loose from any reference to real wealth, circulating in a play of its own. The closure at issue here is continuous with both and parts from both at the same point. Luhmann placed closure in whole systems and treated it as the settled condition of their identity; Baudrillard treated it as the state of an age that had lost its referents. The Treasury case narrows the claim. Closure is taken here as a property of particular devices, present in degrees, reached at a point and capable of being undone, as it would be undone the moment an instrument that had circulated as a reference was forced back to settlement against something outside it. It is a quantity to be read off cases rather than a verdict pronounced on the age.

The 2011 episode is useful because it shows the closure at the one point in the system where there is nothing left above to appeal to. The Treasury could not be downgraded toward anything, because nothing stood above it to be downgraded toward. The market registered this by buying. The regulators registered it by writing the downgrade out of their rules. The scale went on being used. The question we instinctively ask of such systems, whether their verdicts are accurate, assumes a position outside them from which the comparison could be made. For a system that has absorbed its own referent, that position is not available, and its absence is not a defect in the system. It is the condition under which the system works at all.