What Cannot Be Hedged

The sense in which I am using the term is that in which the prospect of a European war is uncertain, or the price of copper and the rate of interest twenty years hence, or the obsolescence of a new invention … About these matters there is no scientific basis on which to form any calculable probability whatever. We simply do not know.

— John Maynard Keynes, “The General Theory of Employment” (1937)¹

In its report for 2003, Global Crossing had to explain what it meant by “we.” A new company had received substantially all the assets of the old one on 9 December. Statements about the business before that date referred to the predecessor; statements about it afterward referred to the successor. The telecommunications network figured in both accounts. Holders of the old company's common and preferred shares received no distribution under the reorganization plan.²

Before there were two companies to distinguish, there had been several futures to accommodate. A customer might expect to need more bandwidth without knowing where his traffic would go. A provider building across oceans had to choose routes long before all that demand could be known. Global Crossing's 1999 report described an arrangement between them: its Network Offer allowed a customer to make a commitment over several years while retaining choices about where and when to activate capacity. The company hoped to obtain loyalty and volume. On the same page it described shorter services and rights extending over a system's expected economic life, different ways of placing a customer's future within the network already built or being assembled.³

The customer was buying some time in which to find out. He could undertake to buy capacity without deciding the geography of all his later traffic, and a provider with enough routes could make that postponement worth selling. Combining demands that arrived at different places and times could improve the use of a network; a customer commitment could help justify capacity that would otherwise be too uncertain to supply. Where payments arrived in advance, they could also help finance it. Both parties might gain room to act from an agreement that neither could obtain alone. Whether this particular offer produced such gains, and at what margins, is more than the report tells us.

The cables' fixed geography helped make a flexible service possible. It also left the supplier with work that a customer need not want to own: assembling the network, keeping capacity available, financing the interval before enough of it was paid for. The customer who wished to defer a choice of route had no corresponding reason to wish the cable's construction deferred. Some commitments enable others to postpone their decisions, and the resulting service may be valuable precisely because it spares its users the need to make the supplier's forecast. Yet a willingness to pay for that service says little, by itself, about how much the supplier should spend to provide it. In the same 1999 report, Global Crossing gave its undersea systems a design life generally of twenty-five years and warned that their economic lives were expected to be shorter. It warned of falling transmission prices and the need for substantial additional finance. These distinctions were available while the money could still be committed differently.⁴

Even the seemingly modest question of when revenue had been earned depended on what had been promised. Global Crossing’s 2003 accounting policy described deferring revenue on flexible bandwidth products and subsea circuits supplied as part of service offerings. Cash received before the earnings process was complete remained deferred revenue.⁵⁶ Money could thus be available for use while years of performance still lay ahead. A service might be useful, properly accounted for, and well priced, yet its provider would still have to survive long enough to deliver it. Advance payment can help make that survival possible. It does not make the remaining work disappear.

Global Crossing's later account describes expansion expected to draw funding from capacity sales on the completed network or from asset sales. Then demand failed to keep pace with added capacity, transmission prices fell, and access to capital contracted. In January 2002 the company entered bankruptcy proceedings.² This is management's explanation, and the accounts themselves were restated; reported pre-collapse revenue cannot simply be read as an uncontested measure of customers' final demand.⁷ Nor was the network a single stock of company-owned cable. It joined owned assets to leases and rights obtained from other providers. The business depended on relations among claims as well as on connections among places. When those relations became doubtful, a customer could have a working service and still need to consider how he would replace it.

VIA NET.WORKS described just that situation in its filing for the quarter ended September 2002. Global Crossing was continuing service on the transatlantic rings where VIA held capacity rights, but VIA could not guarantee what the prospective new owners would do. Existing peering and transit arrangements from its European backbone offered a fallback for traffic to the United States. Depending on them for all its transatlantic traffic might, however, cost more and provide poorer service.⁸ The filing records a contingency, not an interruption. VIA was still using the service and had another way to carry traffic if necessary; neither observation made the existing arrangement dispensable on equal terms. Its capacity rights need not have been acquired under the earlier Network Offer. Here the customer's practical freedom to leave depended on other networks, prior arrangements, and a willingness to bear possible deterioration. The outside option was already a piece of infrastructure.

For the supplier, relief came through a different process. The debtors continued operating within the U.S. and related Bermuda proceedings, while the reorganization allowed obligations to be altered and the business to pass into new ownership. Many contracts and leases were rejected, reducing future obligations. The successor assumed specified liabilities, including obligations under agreements that were retained. Creditors did not all receive the same settlement: certain claims were paid or reinstated, while other creditors received combinations of cash, new equity, and interests in a liquidating trust. The old common and preferred shareholders received nothing under the plan.⁹ A creditor receiving an interest in the successor held a different claim; the shareholder left without a distribution had not acquired flexibility. The business could proceed with fewer obligations because a process had determined how existing claims would be treated. This helped preserve operations as well as change what their proceeds would have to support. It gave the continuing business room to act which cannot be credited, without distinction, to the people who had financed the old one.

The continued usefulness of the network cannot settle whether the original investment was sound. It does make the error easy to forgive in retrospect. People did need to communicate; there were cables to carry their traffic; a business could be made of serving them. An investor, however, had needed more than the broad expectation that communication would grow. He had paid for a claim whose return depended on prices, volumes, costs, the time at which cash became available, and the obligations standing ahead of him. Later use of the network does not restore those terms. The thesis can be correct and the trade lose money. But sometimes what was correct was a much smaller thesis than the one for which the money was paid.

---

Waiting might have improved a particular decision. More could have been learned before expenditure became irrecoverable, although the opportunity to learn had to be kept open somehow: by retaining a site, renewing a permission, holding together a workforce, or persuading a customer to remain. A rival might enter in the meantime. Revenue would be forgone. The calculation has no general preference for patience; under some conditions it gives good reasons to build. A network has to exist before its users can discover all the traffic they will put through it, and an installation may require a period of imperfect operation before it can produce reliably. An account that recognizes prudence only in the person who waits for others to incur these costs has overlooked much of what investment does.¹⁰

To calculate the value of waiting, however, we have to say something about the future in which the later decision will be made. If possible prices and their probabilities can be specified, one can ask how much the opportunity is worth. That is a considerable achievement. Keynes was concerned with expectations for which this specification itself lacked a scientific basis. Giving a model another branch does not ensure that the branch describes what will matter, while refusing to model anything would leave useful distinctions among known exposures unexplored. A firm may have a fairly definite sensitivity to a quoted fuel price, a disputable forecast of next year's orders, and only a poor description of how a new invention could change its business. These difficulties coexist in the same investment. Calculation can improve a decision without turning all three into the first kind of problem.

Even when a change can be foreseen, the agreement may leave its consequences unsettled. Describing every permitted adaptation can cost more than the parties are willing to spend; some decisions are left to the owner of an asset, others reserved to a customer or constrained by a lender. A specialized installation may be worth more in this relationship than outside it, which gives both parties reason to continue and reason to bargain over the terms. Neither dependence nor incomplete agreement is necessarily a defect. They may accompany work that is worth doing together, with details better decided when the need arises. But the person who can vary output may lack permission to sell the plant, and the person who has paid for service may have no authority to reorganize its production. Before assigning a value to an opportunity to change, we have to establish who holds it and what exercising it would require.¹¹

A hedge addresses part of this difficulty by offsetting an exposure. It may use a standard market instrument or an existing operation whose receipts rise with the input cost being protected; no specially negotiated agreement is required in every instance. Protection can preserve the funds needed for a worthwhile project when raising replacement finance would be costly.¹² It can also require funds along the way. In 2008, after its emergence from bankruptcy, Calpine reported that commodity-market movements had increased collateral requirements and margin calls in the second quarter, followed by a decline in the third. Liens on assets and credit facilities helped support its procurement and hedging arrangements.¹³ The company was providing for a demand on liquidity that could arise before the earnings being protected arrived. This is no evidence that the hedge failed, still less an explanation of its earlier bankruptcy. It is a reason to examine how the protection will be maintained during the period for which it is needed.

Selling an investment does something else. A shareholder accepts an available price and relinquishes the future attached to the share; if expectations have already deteriorated, the sale may realize the loss he hoped to avoid. Refinancing may give a company more time, provided someone will supply funds on terms it can meet. Reducing production can conserve cash without recovering what was spent on construction. Each can be a sensible response, but none is made a hedge merely by leaving its author better able to endure what follows. Nor does the age of the machinery tell us which response is available. A plant can operate for decades while its market value changes this afternoon, its owner faces a payment next month, and its customer retains a contract extending beyond both.

The attraction of a scarce complement, then, lies partly in arrangements which may change well before the complement wears out. Suppose a computational provider can redirect work among several sites. Faster calculation might improve the allocation within the rights it already holds, reducing idle capacity or helping it respond to a customer's request. If relocation is technically possible and permitted by its contracts, the provider may also become less dependent on a particular site. An owner expecting this freedom to be exercised may ask for minimum purchases or compensation for cancellation; the provider may willingly accept such terms to secure capacity it could not otherwise obtain. More capable computation can therefore coexist with a substantial commitment to keep buying. Whether speed changes the bargaining relation, and by how much, would require evidence distinguishing its effect from those rights, commitments, and financing arrangements. The history of an undersea network does not supply that evidence for us.

An alternative use can matter even when no customer is prepared to promise much. The Joule Standard's comparison with proof-of-work mining applies only where the operator can actually convert the constrained resources, taking account of the cost and risk of both uses. A mining quotation offers little protection to someone who cannot make the switch. VIA's backup routes raised a further difficulty: even an available alternative could leave its user paying more for poorer service.

There are more substantial reasons the investment account could disappoint. A cognitive advantage may persist despite the diffusion of particular techniques, leaving its owner with earnings that a prediction of rapid commoditization missed. Cheaper actuation could reduce the scarcity on which another owner relied. Or the prospect of a bottleneck might draw enough capital to relieve it. A gate can move, widen, or attract a queue of people building another gate. Even if it remains scarce, an investor can pay more for it than its future earnings justify. That last possibility lacks the interest of a technical reversal, which may be why discussions of technological change give it less attention than it deserves. There need be no error about the importance of the industry. The purchase price can be sufficient.

Perez's distinction between installation and deployment allows more time for the uneven passage from a new capability to its ordinary use, without supplying an absolution for money spent during installation.¹⁴ A supplier may be ready while its customer is still reorganizing; by the time another supplier is ready, its prospective customer may have found something else. In a regulated workflow, the interval between a credible demonstration and audited, insured use at scale is one way of observing this delay, provided the beginning and the conditions of deployment are specified.¹⁵ A persistent shortening would weaken positions whose expected returns depend on a long institutional interval. But financial time need not follow deployment time. Prices may anticipate a change; contractual receipts may continue after it; refinancing may become difficult while everyone still expects eventual success. The investor cannot obtain his remaining time by reading a single stage of the technology's progress.

A large discontinuity could defeat provisions made for gradual change. So could a decline in the price of an unchanged service, if that was the exposure a particular investor had left unprotected. It is tempting to make the grand event the one risk beyond ordinary management, leaving the rest to some combination of selling, waiting, and borrowing. The history here gives little support to that comfort. Some exposures can be reduced through markets or operating choices; others may remain because protection is too expensive, because the relevant market or arrangement is unavailable, or because the contingency is difficult to describe in terms that would make protection effective. It can be difficult even to establish which limitation governs until someone tries to obtain the protection. Uncertainty remains within these ordinary decisions, alongside risks that can be specified and provisions which may work quite well.

None of this makes reversibility the proper ambition of every investment. A promise of continued supply lets another person plan; stable funding makes possible work that cannot repay it immediately. A customer who bought time from Global Crossing could be buying something useful, and a continuing network could serve people who had no part in financing its construction. The difficulty begins when those uses are offered as a sufficient account of what happened to the commitments that supported them. The 2003 report's distinction between predecessor and successor is worth keeping: it obliges us to say whose history we are following. The capability may spread, the service may improve, and the business may continue under new ownership. None of those observations, by itself, tells us what happened to the people whose commitments made the earlier arrangement possible. For that we must keep following the claims after the forecast has ceased to organize them.

Source notes

Footnotes

1. John Maynard Keynes, “The General Theory of Employment,” Quarterly Journal of Economics 51, no. 2 (1937), 209–223, at 214. The ellipsis marks the omission of the clause about private wealth-owners in 1970. Keynes distinguishes the uncertainties in the epigraph from calculable probability; this does not invalidate probabilistic analysis where its assumptions are adequate. Existing manuscript identifier: keynes1937. ↩

2. Global Crossing Limited, Form 10-K for the year ended 31 December 2003, “Introduction,” p. 1; “Competition,” p. 10; “Events Leading Up to the Chapter 11 Reorganization,” p. 12; “Effect of Consummation,” pp. 13–14. The explanation of financing and market conditions is the company's account. The transfer of substantially all assets does not establish the survival of every original asset or returns earned by successor investors. ↩ ↩²

3. Global Crossing Ltd., 1999 Form 10-K, filed 17 March 2000, “Structured Bandwidth Services,” printed p. 4, reproduced in Florida PSC filing 08577-2000, PDF p. 83. The document describes the offer and management's intended loyalty and volume benefits. The customer buying time, possible pooling gains, and possible financing benefit of advance payment are economic interpretations, not findings about a particular contract, payment schedule, or realized margin. ↩

4. Global Crossing, 1999 Form 10-K, “Risk Factors,” printed pp. 18–19, same Florida PSC PDF, pp. 97–98. The filing separates design life from expected economic life. Neither is a measure of the interval over which market value can change. Construction time, useful life, economic obsolescence, contractual term, and repricing are therefore kept distinct; the chapter does not use duration as a common measure of them. ↩

5. John M. Morrissey, Deputy Chief Accountant, SEC, testimony before the House Committee on Financial Services, 21 March 2002, “Telecommunications Capacity Purchase and Sale Agreements,” particularly Steps 1–2 and “Industry Practice.” Under the guidance described, service revenues accrue as performance occurs; operating-lease revenue runs over the term; sales-type treatment requires additional conditions. The testimony discusses Global Crossing's disclosed investigation but does not adjudicate particular contracts. Its “Accounting for Nonmonetary Transactions” section distinguishes exchanges eligible for fair-value treatment from those requiring carrying values and calls for analysis of the facts. Reciprocal transactions are not all declared improper. Industry guidance does not retrospectively classify every Network Offer agreement. ↩

6. Global Crossing, 2003 Form 10-K, Note 3, “Significant Accounting Policies,” “Revenue Recognition,” printed F-15–F-16, especially “Operating Leases.” The company describes flexible bandwidth and subsea service arrangements with deferred revenue; its advance-cash policy is also explicit. This is company-specific reporting policy, not an established finding about each early contract. “Non-Monetary Transactions,” F-16, reports the company's application of later guidance and adjustments. Revenue recognition, cash receipt, demand and profitable fulfillment remain distinct. ↩

7. Global Crossing, 2003 Form 10-K, “Our Network,” pp. 5–6; “Selected Financial Data,” pp. 47–48, including restatements; “Management's Discussion and Analysis,” pp. 49–50 and 55–57. The chapter supplies no complete causal ranking of accounting, demand, overbuilding and finance, and does not infer that premature arrival alone defeated an otherwise sound investment. ↩

8. VIA NET.WORKS, Inc., Form 10-Q, quarter ended September 2002, Exhibit 99.1, “Risk Factors”, printed pp. 4–5, paragraph under “Some companies with which we do business may have financial difficulties…”. Continued service is reported fact; the fallback, possible additional costs and poorer service are the company's conditional assessment. The filing establishes neither an actual interruption nor realized fallback costs, and does not identify VIA's IRU purchase as a Network Offer contract. ↩

9. Global Crossing, 2003 Form 10-K, pp. 12–14. These describe the Chapter 11 proceedings, related Bermuda proceedings, contract rejection, transfer, assumed liabilities, and treatment of claims. References to creditors compress differentiated classes, not an equal recovery. The passage concerns powers and distributions exercised through this reorganization, not an option freely exercisable under an original contract. It makes no general claim that rejection extinguishes every counterparty right. ↩

10. Robert S. Pindyck, “Irreversibility, Uncertainty, and Investment,” NBER Working Paper 3307 (1990), printed pp. 1–6; published in Journal of Economic Literature 29 (1991), 1110–1148. The initial example specifies future prices and probabilities. Waiting has value and costs within stated assumptions; the model does not supply the probabilities whose calculability Keynes questions. The chapter's illustrations of retaining sites, permissions, personnel and customers apply the problem without attributing them to Global Crossing's actual decisions. ↩

11. Oliver E. Williamson, “Transaction-Cost Economics: The Governance of Contractual Relations,” Journal of Law and Economics 22 (1979), 233–261, especially 239–242; Sanford J. Grossman and Oliver D. Hart, “The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration,” Journal of Political Economy 94 (1986), 691–719, especially 693–696 and the abstract. Specificity concerns value within a particular relationship, not physical durability alone. Rights can be costly to enumerate even where events are foreseeable. Grossman and Hart assume costless ex post recontracting in their baseline model; their result does not establish the renegotiation costs of the historical case. Waiting, specificity and residual control are serious antecedents of the chapter's synthesis, not successive discoveries claimed here. ↩

12. Kenneth A. Froot, David S. Scharfstein, and Jeremy C. Stein, “Risk Management: Coordinating Corporate Investment and Financing Policies,” NBER Working Paper 4084 (1992), abstract and printed pp. 1–3; published in Journal of Finance 48 (1993), 1629–1658. The financing argument depends on external funds being more costly than internal funds; it does not show that every hedge adds value. Standard instruments or offsetting operations can address specified exposures while leaving price mismatch, quantity, credit or liquidity risks. The chapter offers no portfolio recommendation or general impossibility theorem about hedging. ↩

13. Calpine Corporation, Form 10-K for 2008, “Liquidity and Capital Resources,” pp. 54–56; emergence from bankruptcy on 31 January 2008, p. 53. The second- and third-quarter collateral movements occurred after emergence. The account does not claim that Calpine failed to meet its calls, or that this episode caused its 2005 bankruptcy. The market heat rate mentioned in the filing is a price relation, not plant efficiency. ↩

14. Carlota Perez, Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages (Edward Elgar, 2002), introduction, pp. xvii–xviii; chapter 4. The installation/deployment distinction is retained; the limits on retrospective investment vindication are the chapter's inference. Existing manuscript identifier: perez2002. ↩

15. Demo-to-deployment lag denotes the interval between a credible demonstration and audited, insured, or otherwise institutionally recognized deployment in a defined workflow. The starting event and deployment threshold must be declared. The body uses the narrower regulated-workflow example; neither definition supplies a universal threshold or transition law. ↩