Printer Friendly

Risk-bearing contracts for space enterprises.

Risk-Bearing Contracts for Space Enterprises


Enterprises relying on the installation of hardware outside Earth's atmosphere face considerable risk. In the future, such enterprises might include research and manufacturing activities on an orbiting platform. For now, the main commercial exploitation of space lies in the use of satellites for communication, land or sea surveying, and for navigation. Apart from the usual commercial risks, these enterprises face the high probability that hardware will be destroyed on launch or will malfunction in the largely inaccessible environment of outer space. (1)

With some lag, the development of commercial space enterprises was accompanied by the emergency of an infant space insurance industry. Cover has been provided for ground, launch, "in orbit" and third party risks. The structure of the insurance market has been influenced by the prevailing contracting practices in the launch and satellite manufacturing sectors. For example, manufacturers typically deliver satellites to their customers on the ground, and launch contractors do not typically assume liability for damage to their cargoes. Thus, the owner or operator of the satellite is left bearing the risk for launch and equipment failure that not only removes space-based hardware, but leaves ground based facilities (e.g. receiving and transmitting stations) unused.

Cover for satellite owners and operators has been available, but a spate of losses in the early to mid 1980s led to skyrocketing rates and, eventually, to the virtual disappearance of this insurance market. Though there were signs of a revival in 1987 and 1988, the market is tentative and fragile. Failures in all Western launch systems in 1986 (especially Challenger which, in fact, did not have insured cargo) apparently has caused all major parties to revise their loss probability estimates upwards. Though there have been successfully launches of all major launch systems since then, the confused loss experience leaves room for considerable diversity in estimating loss distributions.

This article examines present risk-bearing arrangements in which satellite manufacture and launch contracts usually assign risk to operators/owners who then pool risks through conventional insurance contracts. It shows how such arrangements may have contributed to the space insurance "crisis." I then examine alternative risk-bearing arrangements to see whether they more appropriately match the risk-bearing capacities of the parties and whether they provide more appropriate incentives for loss prevention.

II. Insurability of Satellite Risks

Recently, Berliner (1982) described a set of conditions which lead to a well functioning insurance market. These are not preconditions, since the violation of one of these conditions does not preclude the existence of a market. The better the performance against these conditions, the more likely it is that markets will flourish. Some of Berliner's conditions do not pose unusual problems for satellite insurance. For example, Berliner suggests that the loss should be measurable [see Kleindorfer (1986) and Kunreuther (1987)]. An example of a major violation may arise in insuring liability for pollution where a spillage of hazardous materials may have environment and health consequences which are not understood and which may take tens, hundreds, or thousands of years to work themselves out. The determination of awards for such damasge by courts is a highly subjective process whigh might go through many appeals and reversals, taking many years for a final settlement. In contrast, the loss of a satellite is fairly simple to value. The loss is a distinct event and establishing a replacement value is not so daunting a prospect. It is not claimed that there is no problem measuring loss, but simply that the problem is unlikely to be a major stumbling block when negotiating insurance contracts.

Other conditions identified by Berliner have counterparts in non-insurance contracts such as those for the sale of goods, capital or labor. Contract designs often reflect the joint desire for an efficient allocation of risk across parties and efficient incentives for use of scarce resources. Of course, the raison d'etre of insurance is to allocate risk to the most efficient risk bearer. This is best achieved if there is a large number of policies in the insurance pool and if these policies exhibit low correlation. The re-allocation of risk affects incentives for investment in safety and loss prevention. Where such moral hazard is severe and cannot be redressed by price and related incentives in the insurance contract, fewer insurance contracts will be completed and, in the extreme, an insurance market may fail to emerge.

Preconditions for Diversification

Satellite insurance contracts are usually negotiated well ahead of the proposed launch date. An investment in a satellite follows a distinct sequence which usually takes two or three years from the investment decision to the presence of a functioning system with the satellite in orbit. The satellite owner or operator may well wish to define the parameters of the investment decision, particularly the level of risk, at the time of decision. This generates a demand for the insurance contract to be negotiated and conditions fixed well before the proposed launch. At issue, is whether such contracts are compatible with an insurance pool in which the insurer is able to charge adequate premiums and to diversify risk.

The temporal sequence is defined on a time scale from 0 to 1. At time zero the contract is negotiated and time 1 is the proposed launch time. The loss is revealed at the time of launch. This is a simplification since a period of post-launch testing is required to see whether the satellite functions in its space environment after the rigors of launch. The insurer forms expectations about the loss distribution at time 0 and these form the basis for setting a premium. If the insurance pool is to function well, the insurer must be able to predict its average payout within narrow limits, [see Cummins, (1974)]. This requirement is represented in the following form:

[Mathematical Expression Omitted]

where VAR[.] is the variance [L.sub.i] is the actual loss to policyholder "i" n is the number of policyholders E(.) is the expected value based on information available at time t

Consider the insurer's expectation of loss immediately before the launch date, i.e. time 1. I assume that the revealed loss deviates from the pre-launch expectation by a random error, e, and that the expected value of this error at the time immediately before launch is zero. (2)

[Mathematical Expression Omitted]

The error term, [e.sub.i], is a forecast error. It can be thought of as the insurer's static risk since it reflects the fixed technology, or more precisely the underwriter's knowledge of that technology, immediately before launch. However, insurance contracts are usually written months, or even years, before the proposed launch date. Time zero refers to the time the contract is written. The insurer is assumed to form some expectation concerning the loss at time zero. Now, it is quite possible that new information will become available to the insurer after the contract is written which will cause a revision in the underwriter's time zero loss expectation. For example, the insurer might use past experience to predict the future loss distribution in a Bayesian process. If there were few previous launches, any new launches will substantially increase the sample size and could lead to a significant revision of the estimated loss distribution. New information might reveal new technical problems that were not known to exist or were underestimated (e.g., problems with "O" rings in the shuttle's solid boosters) or could reveal redesign features that mitigate previously known problems. With a flow of such new information, the insurer might well hold very different expectations about the probability of loss on the eve of the launch than where held when the insurance contract was originally written. The longer the time between the original contract and the launch, the greater the accumulation of new information and the greater the potential for a revision in loss expectations. Thus, at the time the contract is written, the expected loss, denoted by the underscript "O", differs from the pre-launch, time 1, expectation by some revision factor "u" which might be described as the forecast revision: (3)

[Mathematical Expression Omitted]

Combining (2) and (3) gives:

[Mathematical Expression Omitted]

There is assumed to be no correlation between the original forecast error, e, and the forecast revision, u. Now one can see whether the requirement for predicting the aggregate loss experience is met. For simplicity, assume that the loss distributions facing the operators are identical. Taking the variance of (1) and noting that losses are as shown in (4), gives:

[Mathematical Expression Omitted]

Diversification Problems in the Satellite Insurance Market

Equation 5 reveals three potential weaknesses in the space insurance market. The extent to which a pool is able to diversify risk depends on: (1) the number of exposures insured, "n", (2) the covariances in the forecast errors, COV([e.sub.i];[e.sub.j]), and (3) the covariances in the changing information base, COV([u.sub.i];[u.sub.j]). Ceteris paribus, the smaller is "n", the larger is COV([u.sub.i];[u.sub.j]), the larger is COV([e.sub.i];[e.sub.j]), the greater is the riskiness of the insurance pool. In satellite insurance, each of these effects is important.

First, satellite insurance pools are small. Figure 1 shows the number of insured commercial spacecraft launches from 1965 through 1986. In recent years, the number of insured launches per year was 20 or less. Moreover, these were not all covered by all underwriters. Thus, each underwriter has carried only a handful of coverages in any year. This is an insufficient base from which to diversify risk effectively. However, there are two features of insurance pools that mitigate the adverse effects of a small sample of exposures in the insurance pool. The first mitigating feature of small size is that gains from diversification encounter diminishing returns. The risk pattern implied by equation 5 is depicted in Figure 2. The larger the number of exposures, the smaller the relative risk. If the covariance terms are non-zero there will be a limit, below which further diversification will not reduce risk. This limit is given by the final term in the right hand side of equation 5. Thus even very large insurance pools would be left with a substantial degree of relative risk. The second mitigating feature is that satellite insurers rarely insure only this line of business. Most are aviation underwriters and many also insure other lines. The financial performance of the insurer is determined by the market, not by the level of risk on any single line of business, but on its overall portfolio. Thus the satellite risks can be, and are, pooled with all other business. This "cross line pooling" can dramatically reduce the overall risk to the firm if the cross line correlations are low.

The second constraining feature on risk reduction relates to the covariances in the forecast errors. Covariances in these forecast errors may arise even if the errors are unbiased (i.e., E([e.sub.i]) = 0 for all i as assumed earlier) simply because the events themselves are not independent. The obvious example arises with the delivery of two or more spacecraft on a single launch vehicle. A catastrophic launch failure will simultaneously destroy all cargo thus giving rise to total loss claims for each of the insured satellites. The effect of such correlations again is depicted in Figure 2. Correlation in forecast errors, together with correlations in the information base discussed below, jointly determine the amount of undiversifiable risk in the pool. If these correlations are high, then increasing the number of policies will have only a limited effect in reducing risk.

The third feature determining the risk in the insurance pool is the covariance structure in the changing information base, COV([u.sub.i];[u.sub.j]). Covariances may arise in several ways, [Several of these issues are discussed in Goudge (1987)].

1. First, the sample size changes over time. The insurer's estimates of future loss probabilities are likely to be based partly on past loss statistics and partly on other information. The probability that a satellite will be successfully launched and deployed in the correct orbit depends largely on the particular launch system used. Past statistics for any given launch system used in a particular configuration, and for the particular function of the proposed launch, are likely to be very small. Table 1 shows comparative failure rates for different systems with geostationary launches separated out. The sample sizes for most systems are small. Yet bear in mind that these samples are misleading. For example, early Arianes are not the same as later Arianes. The system has evolved to larger and more sophisticated vehicles, yet has had time to sort out its teething problems. Thus the effective sample size, say for Ariane 4, is much smaller than indicated in the Table. With such small samples, any new launches may considerably change the information base. For example, suppose that a system has had 5 launches with 1 failure. A new launch could change the failure rate from 20 percent to 33.3 percent if it failed, or to 16.67 percent if it succeeded. While underwriters would not rely solely on statistical information, it becomes clear that a small increase in the sample could cause a significant revision of the estimated loss distribution for all policies committed on this launch system.

2. A second way in which covariances could arise in the information base relates to new technical information. A design change which is intended to increase reliability, give new functions to the launch system, increase its payload, etc. may cause the underwriters to revise their estimates of the parameters of the loss distribution for launches on this system. The effect may be favorable or unfavorable. But the important point is that all committed policies on this system are affected in the same way. As with the effects of increase in the sample, this risk is undiversifiable.

3. The third mechanism relates to new information on organizational procedures that affect quality control or safety. For example, pressure on the NASA to speed up its launch schedule, or budget cutting for the agency, could affect both quality control and its choice of safety margins. Over a given period, this environment would change the loss probabilities for all launches by the agency. Any organizational or procedural changes of this sort, that become apparent to underwriters after policies have been committed, create an undiversifiable risk similar to that from new statistical or technical information.

The problem with new information, i.e. information that becomes available to the underwriter after the policy has been written, is that it cannot be used in determining the terms of cover or the premium to be charged. New information will reveal that the premium was too high or too low. If this error is uncorrelated across policies, there is little problem; it will simply diversify out in the insurance pool. The problem arises if the error is correlated across policies as shown in equation 5. With such correlation, the underwriter will find that all related policies were underpriced or overpriced. This risk is not diversifiable.

Moral Hazard with Current Contracting Provisions

The chances of launch failure or spacecraft failure are not necessarily independent of the risk contracting arrangements. Many types of contracts transfer risk between the parties. In an insurance contract, the transfer of risk to the insurer relaxes incentives for the policyholder to behave safely or to invest in loss prevention. The cost imposed by such behavior is moral hazard. The extent of moral hazard will depend on the cost to the insurer of monitoring the policyholder's actions. If these actions are easily observed, the insurer will rationally include incentives that condition the premium and/or the terms of cover on the policyholder's action [see Marshall (1979), Rubinstein and Yaari (1982)]. For example, in a property insurance policy, the insurer may offer a sizeable premium reduction if sprinklers are installed.

Though moral hazard is well documented in insurance contracts, similar problems arise with other types of contracts. For example, if the manufacturer of a product provides a comprehensive guarantee on its performance, the purchaser may be disinclined to exercise costly and inconvenient care when using the item. This type of contract also encounters moral hazard. The wisdom of a comprehensive guarantee might be questioned. The comprehensive guarantee assigns the risk of failure of the product to the manufacturer. Without a guarantee, and in the absence of protective consumer laws, the risk would fall on the consumer. Which allocation of risk is preferable?

If risk can only be assigned to one party, and if both parties are risk neutral or can buy actuarially fair insurance, the socially optimal solution is to choose the assignment that has lower total costs. If there were no transaction costs (i.e., monitoring costs) to allocation of risk would be irrelevant. This would be a straightforward example of the Coase theorem [see Coase, (1962)]. In practice, the transaction costs can be significant. Thus the resolution requires that risk be assigned to the party that can reduce the total cost of failures at lower cost.

Of course there may be a better solution still. A limited guarantee may make the manufacturer responsible for some failures (e.g., those due to product design or to poor quality control in the manufacturing process) and for the consumer to be responsible for other failures (e.g., those due to poor maintenance or to improper use of the product). This allocation can be explained with the same principle of minimizing the total cost of failure including prevention and monitoring costs.

The typical assignment of risk in the broadly defined "satellite industry" already has been characterized. Manufacturers usually deliver on the ground to the operator and the launch contractor does not assume responsibility for loss of its cargo (other than possibly a limited insurance for the costs of a subsequent launch). Thus the operator is left bearing the risk of launch or "in orbit" failure. This risk is sometimes passed to an insurer.

Now consider which parties have control over the risk of failure of the launch or of the operation of the spacecraft. Failure may arise because the satellite is poorly designed or manufactured; because it is destroyed on launch, it is damaged in the harsh launching environment or it is delivered to an incorrect orbit; or because it is operated or deployed ineffectively by the operator in such a way as to affect its life and performance; e.g. excessive and unnecessary corrections to its attitude and orbit will lead to early exhaustion of the fuel.

The most likely causes of loss are the first two listed above. The design and production is in the hands of the manufacturer. The operator will have no direct control over the standard of workmanship in the production of the satellite. The operator may exert an indirect control over design and production if it can monitor the behavior of the manufacturer. But monitoring such a technically complex process is likely to be very costly. Thus, it would seem that a strong prima facia case exists that total failure costs would be lower by assigning the cost of failure due to design/production defects to the manufacturers.

Similar considerations apply to failures stemming from the launch service. Launch agencies such as NASA directly control the launch operation and indirectly control the quality of the launch hardware and software through a combination of design specification, performance incentives to subcontractors, monitoring of subcontractors, and direct testing of systems. Again, the operator has little direct control and monitoring costs are likely to be extremely high. In fact, monitoring costs can become infinite in theory if the agency is protecting technical secrets, perhaps for military purposes. Again, a strong prima facia efficiency case can be made for allocating the costs of launch failure to the launch contractor.

There are other complicating features of the commercial space marketplace. Suppose that this market were competitive, the actors had long time horizons, and satellite operators made repeated contracts with the manufacturers and launch contractor. Even if risks were assigned to the operator, the manufacturer and launch contractor would still select socially optimal levels of safety. If they did not, they would certainly lose their existing customers, and probably their potential customers, to rival firms. But these conditions do not hold. The satellite manufacturing industry is highly concentrated and launch services have largely been provided by quasi government bodies (NASA and Ariane). This is not to say that competition does not exist; rather that price signals are likely to be distorted by political (and other) noise. The 1986 decision to spin off commercial launches from NASA to private contractors may lead to significant changes, (see "Rocket Makers Tackle a Private Launch Puzzle", Wall Street Journal, p. 35, 8.4.87.)

How does insurance fit into this picture? Ironically, if it is argued that the operator has little control over the probability or severity of satellite loss, an insurance contract to indemnify the operator will face little moral hazard. While it is true that insurance will reduce the incentive for the operator to monitor the manufacturer and the launch contractor, it is likely that high monitoring costs would have deterred this activity anyway. In fact, it is possible that insurers covering many exposures would reap scale economies in monitoring activities of the manufacturer and launch contractor. If so, the cost of moral hazard in the insurance contract would be negative. If insurers were able to better perform this monitoring function, this would mitigate the adverse incentives in the satellite purchase and launch contracts. Nevertheless, the monitoring burden is still there and may still be costly.

III. Contracting Alternatives for the Commercial Space Industry

The preceding section presented two broad criteria on which contract design might be examined. The first was concerned with the sharing of risk between parties: more specifically, the functioning of insurance markets. In a well functioning insurance market, the insurer is able to bear risk at lower cost than can a policyholder since the insurer can diversify over a large number of exposures. The satellite insurance market did not reveal such a clear comparative advantage for the insurer in risk bearing. This feature is important in understanding the difficulties in this market.

The Trade Off Between Risk Sharing and Moral Hazard in Contract Design

An insurance contract is specifically designed to shift risk between the parties. But other contracts might also, less directly, transfer risk. For example, a procurement contract for some raw material will determine who bears the risk of changes in the market price in the future. A labor contract will often determine the conditions under which employees may be laid off. Other things being equal, it is efficient for the contract to assign risk to the party who can bear that risk at least cost. Thus, many labor contracts provide some degree of job protection to the employees since the employer is able to bear that risk at lower cost.

The second criterion for examining contract design is related to moral hazard. In the absence of risk preferences, the contract that presents the lowest cost in terms of transactions (e.g., monitoring) and resource misallocation (e.g., moral hazard) is preferred. But it is clear that a comprehensive examination of contracting in the satellite industry requires that both allocational efficiency and risk sharing be addressed. Alternative contractual arrangements will be examined. The first, and least radical, considers only a change in the design of the insurance contracts sold to operators.

A Redesign of Insurance Contracts Sold to Operators

The typical contracting arrangements of the satellite industry are as follows. Operators contract for the purchase and launch of satellites but are left with the cost of launch or "in orbit" failure. These risks are insurable. The simplest revision of the current contracting pattern lies in a redesign of the insurance contracts while retaining the assignment of risk to the satellite operator in the satellite purchase and launch contracts.

To see how insurance contracts might be improved, refer back to equations 2, 3 and 4. Consider an insurance contract that is written at time 0. At that time, the underwriter's expectation of loss is:

[Mathematical Expression Omitted]

If the premium is set at time 0 and is a constant, it cannot by definition include the information contained in the forecast error, e, and in the forecast revision, u. An example of a premium set at time 0 is shown in equation 6 with R being some risk premium.

[Mathematical Expression Omitted]

Now assume that the premium is set at a later date. One possibility is that the premium is set when u is revealed but before the actual launch. This premium is shown in equation 7.

[Mathematical Expression Omitted]

A more radical possibility is that the premium is set after the loss is revealed. Clearly, there would be no transfer of risk if the premium was equal to the revealed loss. But suppose that the premium was set equal to the expected loss at time 0 (or this could be done at time 1) plus an adjustment that allocated any difference between revealed aggregate premiums and aggregate losses over all policyholders insured in a given period. This is labelled a retroactive premium, thus the underscript r:

[Mathematical Expression Omitted]

With such premiums, total premiums will equal total losses ex post. Thus, since homogeneity is assumed, each insured ultimately pays a premium equal to the average loss. (4)

It if fairly routine to show the risk in the insurance pool with each of these three premiums. The premium defined by equation 6 exposes the insurance pool to risk arising from the new information, u, and from the forecast errors, e. Thus the variance of the insurer's underwriting performance is, as shown in equation 5,

[Mathematical Expression Omitted]

If premiums are as shown in Equation 7, the insurance pool no longer is faced with the risk inherent in u. The risk is passed back to policyholders by delaying the time at which the premium is charged. Thus the pool is left only with the risk arising in e. The variance of the underwriting results will be:

[Mathematical Expression Omitted]

Delaying the date at which the premium is fixed removes part of the problem of diversification within the insurance pool. But this process can be taken further with the retroactive premium described in equation 8. Under this premium, the total losses equal total premiums ex post leaving a riskless insurance pool.

Premiums are able only to impound information available at the time the premium is fixed. By delaying the setting of premiums, more information may be impounded. Such delays may pass the risk inherent in that information back to the policyholder. Two obvious questions follow: does this retroactivity in the setting of premiums improve the welfare of the parties and, if so, how may such retroactivity be achieved?

In addressing the first question, it is important to note that the premium schedules described in equations 7 and 8 may be used to decompose risk into that which is diversifiable and that which is not. The idea is to permit the insurance pool to absorb the diversifiable component of risk undiversifiable component (or at least part of this component) with the policyholder. Marshall, (1974) showed that, in a market of individual insurance buyers, a retroactive premium schedule (equation 8) is Pareto superior to a simple prepaid premium (equation 6) if there is undiversifiable risk. This result has been extended by Doherty and Dionne (1987). The intuition behind this result is that a risk premium will be charged for risk that is not diversifiable within the market. With a simple insurance premium (equation 6) this risk premium acts as a deterrent to the purchase of insurance. With the retroactive premium, (equation 8), the risk is decomposed and the undiversifiable risk is insured without a risk premium. A separate contract can be written to share the undiversifiable risk according to the relative risk tolerances of the parties. The premium schedule described by equation 7 is a half-way measure. These results have been applied to corporate insurance decisions in Doherty (1987).

Given the argument that retroactive premium schedules are Pareto superior to simple prepaid premiums, how may such contracts be designed for satellite risks? One possibility is simply to shorten the commitment period on satellite policies. Cover has typically been offered, and the premium fixed, many months, perhaps two or three years, before the proposed launch date. The cumulative effective of new information picked up in the "u" term may be great and this risk may not be diversifiable. By shortening the commitment period the risk is partly passed back to the insured. A less radical measure is to commit cover, but to defer the assessment of the premium. There is some sign that the market is now responding in this fashion. Prior to 1986, commitments of three years were not uncommon. From 1986 through 1987, not only did rates soar, but commitments were shortened to as little as three months. The softening of the market in the second half of 1987 has seen commitment creep up to a year; but still nowhere near the levels of earlier markets.

The second possibility is more dramatic. Completely retroactive premiums, (equation 8) may be achieved in one of two ways. First, insurers can charge an advanced premium but offer a retroactive dividend or assessment which is calculated according to the second term in equation 8. Such policies are common in life insurance and are called "participating policies". But, to the author's knowledge, they are not used in satellite insurance. The second way is to arrange insurance on a mutual basis such that the policyholders also own the equity in the insurance pool. This may be done by forming a group captive, a risk retention group, or a mutual insurance company jointly owned by satellite operators and jointly insuring their exposures. Though the idea has been an issue for debate among satellite operators, to the author's knowledge no such pools have in fact been formed.

Comparisons with other insurance markets facing similar diversification problems are illustrative. The liability insurance markets face similar problems of changing "technology" between the time the insurance contract is issued and the resolution of claims. In the liability market, "technology" refers to the legal rules which are evoked to resolve liability. The problem appears to have been a destabilisation of these rules leading to substantive and correlated changes in the underlying loss distributions between issue of the policy and the final resolution of claims, [see Priest (1987), Trebilcock (1987), and Winter (1988)]. One response to this has been to substitute claims made policies for the traditional occurrence forms. This has an effect comparable to the shortening of commitments in satellite insurance, since the insurer reduces the period over which it is exposed to changes in liability rules. But another widespread development in the liability insurance markets, (particularly the products, directors and officers, pollution, and municipal liability markets which were hardest hit by the liability insurance crisis) has been the emergence of "mutual like" insurance pools such as group captives and risk retention groups. One possible reason why such pools have not emerged in satellite insurance relates to the sequencing of risks. The main risk of damage is from launch failure. Consider a sequence of ten launches. If risks are pooled between operators, those launching early would appear to have first claim on the resources of the pool (including any rights of assessment which the pool might have over its members) thereby eroding the value of cover for those coming later. Alternatively, the value of cover can be divorced from the sequencing of launches by making a final determination of claim costs, dividends and assessment only after all launches have been completed. But this imposes a cost on the early launches by delaying claim settlement. Such problems might not be terminal to satellite insurance pools in the long run, but they do offer a possible explanation of why pools have not prevailed in this market when they have proliferated elsewhere.

In summary of this section, there are possibilities for simply redesigning the nature of insurance policies sold to operators who would otherwise bear the risk of satellite loss. These reforms involve the use of retroactive premiums. While it can be demonstrated that these alternatives achieve a preferred allocation of risk, there has been modest progress in this direction. The main innovation is a shortening of the time commitments made on satellite policies. The arguments presented here look only at risk sharing, not at moral hazard. The implicit assumption was that the moral hazard problems arising from insurance contracting alternatives will be addressed next in which both risk sharing and moral hazard factors are important.

A More Complete Re-structuring of Contractual Arrangements

The above discussion suggests that moral hazard might be lessened if the cost of launch failure were to be borne by the launch contractor and the cost of satellite malfunction borne by the manufacturer. This allocation of cost could be achieved by rewriting the launch and sale contracts. Another possibility is that the manufacturer deliver the satellite to the operator "in orbit". To achieve this, the manufacturer must subcontract with a launch contractor for launch services. Such arrangements recently have been assembled for the launch of Aussat 2 and for the launch of two satellites for a British company. For example, the contract for the British satellites was assembled between Hughes Aircraft Co., McDonnell Douglas Astronautics and British Satellite Broadcasting Ltd. Under this agreement, Hughes accepts the risk for the satellite to be delivered to the client, British Satellite, functioning and in orbit. Hughes contracted with McDonnell Douglas for the launches on Delta Rockets, [see Stevenson (1987)]. Assuming that the Hughes, et al. contract assigns to McDonnell Douglas liability for damage to the satellite cargo from the launch operation, then this arrangement appears to satisfy the moral hazard criteria. The second question is whether contracts such as these provide an appropriate allocation of risk between the parties. To pursue this line, consider the conditions that might lead one party to have a comparative advantage over another in bearing risk.

Asset pricing models, such as the capital asset pricing model or the arbitrage pricing model, hypothesize that a risk premium will be allocated to a set of traded cash flows only if the cash flows are correlated with the market portfolio or some priced factors. The logic is that, if risk may be substantially and costlessly eliminated by investors through a process of simple diversification, then it will not be priced by market. Only if risk is not amenable to simple diversification within the capital market (called systematic risk) will it command a risk premium. The narrow implication of these models is that, if a cash flow is traded between two firms and if the cash flow were competitively priced, the value of each firm would be unaffected. Thus it would appear that no firm would have a comparative advantage in bearing risk over another. Implicit in this conclusion is the assumption that any change in the riskiness of a firm's cash flows has no effect on the other moments of its earnings distribution; in particular, any change in the variance of earning does not affect the expected value of earnings. If this assumption is relaxed, a comparative advantage may be established by one firm over another in bearing risk.

Several possibilities arise in which reductions in corporate risk may affect the value of the firm. One example arises from bankruptcy cost, the expected value of which may be increase as the probability of bankruptcy increases. Another example arises if any of those having a claim on corporate earnings are risk averse and are unable to costlessly diversify away that risk. Examples would be employees, managers, or customers who retain some post sale claims (e.g., a service contract) on the firm. An even clearer example stems from taxation. Marginal tax rates typically increase with corporate earnings. This concavity is due both to the progressivity of tax rates and to the effects of tax shields. Thus, from Jensen's inequality, the expected value of taxes will exceed the tax which would be due on the expected value of taxable earnings. The tax resembles a call option on the pre-tax earnings and the value of the tax is positively related to the variance of the underlying cash flow [see Smith and Stultz (1986) and Scholes and Wolfson (1987)].

Given these variance related claims on the firm, any reduction in the variance of pre-tax earnings will increase the value of the firm since it will reduce the expected tax liability and the expected value of labor and bankruptcy costs. Such arguments suggest that the risk of satellite failure be assigned to the party that is able to reduce the variance of its earning at lowest cost.

In the present context it seems plausible that a comparative advantage in risk bearing will arise between the manufacturer of satellites and the operator. The manufacturer typically will produce and sell several, (perhaps many), satellites within a given period. If the manufacturer bears the risk of loss for a number of sales, there will be a pooling of risk across the total sales. Does the operator have such a spread of risk? Certainly most operators have few satellites (with one or two exceptions such as Intelsat) and cannot spread risk effectively over their pool of spacecraft. Of course, it is still possible that the operator could combine the risk of loss of a satellite with corporate earnings streams from non-satellite operations. However, it does seem that a prima facia case exists that the manufacturer will be able to spread satellite risk more effectively than will the operator. (5)

The above arguments suggest that a weak case exists for risk to be allocated to the satellite manufacturer rather than to the operator on the grounds of risk bearing. This case is based on the a priori reasoning that the manufacturer can effectively pool the risk with other satellite sales. The proposed assignment of risk to the manufacturer would be reinforced if it were shown that the tax schedule of the manufacturer is more concave over relevant range than that of the operator.

The remaining issue to be addressed in this section concerns insurance, given an assignment of risk to the manufacturer. Arguments identical to those above can be used to rationalize the demand for insurance to protect the manufacturer against this assumed risk. With a portfolio of "in orbit" delivery contracts, the manufacturer has in place a quasi insurance pool. Without any external purchase of insurance the risk will be spread within this portfolio. On the face of it, the role of insurance has been eroded by assigning risk to the manufacturer. An operator with a single satellite may gain much from insurance with a firm underwriting several similar risks for other operators. It is more difficult to establish that a manufacturer bearing the risk of, say, ten launches is at a comparative disadvantage in risk bearing compared with an insurer writing ten such launches. Of course, the direction of comparative advantage is not settled just by making this point. A more complete case for the irrelevance of insurance with this risk assignment requires that possibilities for tax arbitrage, bankruptcy cost and other factors not favor the transfer of the manufacturer's risk to an external insurer.

IV. Conclusion

Satellite insurance pools covering the risk of the operator face significant undiversifiable risk. This undiversifiable risk and arises largely from the effects of new information between the writing of the insurance contract and the realization of loss. Traditional insurance contracts are inefficient in writing such risks. Revised contract designs include contracts with shorter commitment periods and "mutual-like" arrangements in which the equity of the insurance pool is held by the policyholders. While risk sharing arguments may favor redesigned insurance contracts, both moral hazard and risk sharing arguments may be used to suggest more radical re-design of contracting arrangements. These include the "in orbit" delivery of satellites by the manufacturers and, if tax factors are central, the leasing of satellites to operators. Such contracts shift risk from the operator to the manufacturer who is likely to have a comparative advantage in bearing risk. In addition, "in orbit" delivery creates more direct incentives for design improvement and quality control in the production and launching of satellites.

The conclusions of this article favor a form of contracting that, until recently, has not been observed. Moreover, the contracting choices examined do not evoke a trade-off between incentive efficiency and risk sharing, but suggest (though only weakly) a dominant design on both criteria. If the case for contracting choices such as "in orbit" delivery is so clear cut, why have they not been observed in the market? Several issues are relevant. First, the operation of commercial satellites is a relatively new and a small sample activity. The importance of problems such as undiversifiable risk and the quality control in launch vehicles and spacecraft only appears with the emergence of reasonable sample of operating experience. Second, commercial space activity is not a perfectly competitive market. Manufacture of satellites and launch vehicles is highly concentrated and the control of launch activities has been with government, or quasi government agencies. In such a market, the sanction on inefficient forms of contracting is dulled. Such imperfections recently have been reduced with the spinning off of commercial satellite launches from NASA to commercial contractors. Therefore, it seems significant that two recent examples of the allegedly preferred form of contracting, i.e. "in orbit" delivery, have been observed. But two swallows don't make a summer! The sample of such observations is, as yet, far too small to provide a market test of the economic arguments of this article.

(1) The shuttle gives some access to satellites in low earth orbit and satellites have indeed been repaired and/or recovered, i.e., Pallapa and Westar. However, this is an unusual and extremely costly process. Moreover, no such access is available to high earth orbit at present.

(2) This is analogous to the rational expectations framework used to described insurer loss expectations by Cummins and Outreville (1987) and Doherty and Kang (1988). Their focus is on the cyclical behavior of insurance profits.

(3) The distinction between the forecast error, e, and the forecast revision, u, parallels somewhat the Knightian distinction between risk and uncertainty. The term, u, arises from the underwriter's inability, at time zero, to completely specify the parameters of the distribution from which the loss will ultimately be drawn. In this sense, it represents future uncertainty to the underwriter.

(4) If policyholder loss distributions are not assumed to be identical, then a comparable solution would require that retroactive premiums be related to the time zero expected losses.

(5) An alternative tax rationale for allocation of risk may exist even if all cash flows of two firms are identical ex ante, but the concavities of their tax schedules differ; then the firm with the more concave tax schedule will have a comparative advantage in risk bearing. The greater concavity for one firm may arise if it has different tax shields. Such differences give rise to possibilities for tax arbitrage which are exhausted only when such activity equates marginal tax rates across firms. One well known vehicle for such arbitrage is leasing [see for example, Wolfson (1985)]. To bring this discussion to the present issues, there are no obvious analytic reasons to suppose that either the satellite manufacturer or the operator would face a more favorable tax schedule for bearing risk. The matter is empirical. If it were the case that the manufacturer could bear risk less costly for tax reasons, then the appropriate allocation of risk could be made by an "in orbit" delivery contract as described above, or the tax benefits could be reaped through a lease contract with the operator rather than a straight sale.


[1.] Berliner, B., 1982, Limits of Insurability of Risks, Englewood Cliffs, Prentice Hall.

[2.] Coase, R. H., 1962, "The Problem of Social Cost", Journal of Law and Economics.

[3.] Cummins, J. D., 1974, "Insurer's Risk: A Restatement", Journal of Risk and Insurance, 41: 147-157.

[4.] Cummins, J. D. and F. Outreville, 1987, "An International Analysis of Underwriting Cycles in Property Liability Insurance", Journal of Risk and Insurance, 54: 246-262.

[5.] Doherty, N. A., 1987, "Insurance, Risk Sharing and Incentives for Commercial Use of Space", in Economics and Technology in U.S. Space Policy, ed. M. Macauley, Washington, (Resources for the Future and the National Academy of Engineering.

[6.] Doherty, 1978, "Insurance with Random Premiums", Working Paper, The Wharton School, University of Pennsylvania.

[7.] Doherty, N. A. and H. B. Kang, 1988, "Interest Rates and Insurance Price Cycles" Journal of Banking and Finance, 12: 199-214.

[8.] Doherty and G. Dionne, 1987, "Risk Pooling, Contract Structure and the Organizational Form of Insurance Firms", Working Paper, University of Pennsylvania.

[9.] Goudge, B., 1987, "Proton and Long March -- Are They Insurable?", Space, Vol. 3, #2, May-June.

[10.] Hart, O. and B. Holmstrom, 1985, "The Theory of Contracts" in Bewley (ed), Advances in Economic Theory, Cambridge, Cambridge University Press.

[11.] Johnson and Higgins, 1987, A Report on Spacecraft Insurance, New York.

[12.] Kleindorfer, P., 1986, "Environmental Liability Insurance: Perspectives on the U.S. Insurance Crisis", Seminar fuer Versicherungswissenshaft der Univeristaet Hamburg.

[13.] Kunreuther, H., 1987, "PRoblems and Issues in Environmental Liability Insurance", Geneva Papers in Risk and Insurance, 12: 180-197.

[14.] Marshall, J. M., 1974, "Insurance Theory: Reserves versus Mutuality", Economic Inquiry, 476-492.

[15.] Marshall, J. M., 1976, "Moral Hazard", American Economic Review, 66: 882-890.

[16.] Priest, G. L., 1987, "The Current Insurability Crisis and Modern Tort Law" Yale Law Journal, June, 1521-1590.

[17.] A. Rubinstein and M. Yaari, 1983, "Repeated Insurance Contracts and Moral Hazard" Journal of Economic Theory, 30: 74-97.

[18.] Scholes, M. and M. A. Wolfson, 1987, "Taxes and Organization Theory" Working Paper, Stanford University.

[19.] Smith, C. W. Jr. and R. Stultz, (1985), "The Determinants of Firm's Hedging Policies", Journal of Financial and Quantative Analysis, 20, 391-406.

[20.] Stevenson, R. W., 1987, "Satellite Risks Shiften by Pact" Wall Street Journal, August 9: D2.

[21.] Trebilcock, M. J., 1987, "The Social Insurance-Deterrence Dilemna of Modern North American Tort Law: A Canadian Perspective on the Liability Insurance Crisis", San Diego Law Review, 929-1002.

[22.] "Rocket Makers Tackle a Private Launch Puzzle", 1987, Wall Street Journal, August 4: 35.

[23.] Winter, R. A., 1988, "The Liability Crisis of 1984-1986 and the Economics of Competitive Insurance Markets", Working Paper, Yale Law School.

[24.] Wolfson, M. A., 1985, "Tax, Incentive, and Risk Sharing in the Allocation of Property Rights: The Generalized Lease-or-Buy Problem", Journal of Business, 158-171.

Neil A. Doherty is Professor of Insurance at the Wharton School of the University of Pennsylvania. He wishes to thank Molly Macauley of Resources for the Future and an anonymous JRI referee.
COPYRIGHT 1989 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Doherty, Neil A.
Publication:Journal of Risk and Insurance
Date:Sep 1, 1989
Previous Article:Analysis of productivity at the firm level: an application to life insurers.
Next Article:Interest rates, the emergency fund hypothesis and saving through endowment policies: some empirical evidence for the U.K.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters