Jan 1, 1996
An Examination of Alternative Approaches
The Air Force spends over $1 billion annually on contracted weapon system maintenance. Further, recent proposals by the Commission on Roles and Missions and others have sought to increase the fraction of Air Force and Department of Defense (DoD) weapon system maintenance provided by contractors. In light of the large and potentially growing role of contractors in the weapon system maintenance process, we thought it valuable to examine how the Air Force, specifically, and the DoD, more generally, has and could design contracts for weapon system maintenance.
In Government Contracting Options: A Model and Application, the author suggests that contracts that measure performance by the number of flight hours aircraft operate induce better contractor behavior than contracts that use other performance measures.
Historically, a common approach to weapon system repair and maintenance could be categorized as "per repair" contracts. With a per repair contract, a contractor is compensated on the basis of the number of broken units fixed. For example, one type of per repair contract is a requirements contract. With a requirements contract, a contractor agrees to repair broken items for a specified fee per unit without an up-front specification of how many units will need to be repaired. Another type of per repair contract is a fixed price/fixed quantity contract, in which both the price per unit and the number of units are specified ahead of time. Still another per repair contract variation is a time and materials contract. With a time and materials contract, the contractor is assured of being reimbursed for materials used plus an hourly rate per labor hour necessary. Although there is no formal link between the number of units repaired and payment under a time and materials contract, there is a clear correlation between the number of units repaired and the time and materials required.
There has been some experimentation in the DoD with a different contract form we label "availability-oriented." With an availability-oriented contract, the contractor is compensated for keeping a weapon system available irrespective of how many repairs are actually undertaken. For example, Serv-Air Inc. has a contract with the Air Force for repair of the C-21, the military version of a Lear Jet. Serv-Air is compensated on the basis of the number of C-21 flight hours the Air Force receives, not the number of repairs Serv-Air performs (although Serv-Air does receive separate fees for unusual, high-cost "over and above" maintenance and repair). Table 1 contrasts some important provisions of the C-21 contract with a typical per repair government contracting approach. Also, the Navy has a contract with Litton Industries for repair of the LN-15C inertial navigation unit, in which Litton receives a payment per LN-15C flight hour while guaranteeing the Navy a specified availability level.
|Contract Attribute||A Typical Per Repair Approach||C-21 Contract|
|Contractor payment||Based on number of repairs performed||Based on the actual number of hours flown|
|Minimum availability||NA||85% Mission Capable Rate|
|Spares||Government provided||Contractor provided|
|"Over and above" maintenance||Contractor cost plus fee||Fixed prices for "over and above" maintenance and repair|
|Contract duration||One-year contract with one or two government-held one-year options||One-year contract with nine government- held one-year options|
|Safety||Contractor follows government safety rules, government liable for damaged assets||Contractor indemnifies for assets damaged while in contractor possession|
Source: Air Force Solicitation Number F34601-94-R-0032.
Obviously, an availability-oriented contract is quite different from a per repair contract. Are there reasons to prefer one approach to another? Are there other repair contracting approaches the Air Force and DoD should consider?
To evaluate these questions, we built a model of a repair contractor that maximizes its own well-being. We then simulated how such a contractor would respond to different sorts of contracts the government might offer.
Figure 1 sketches the model we developed. In the model, the contractor chooses a level of repair capacity, a repair quality level, and a repair speed. The contractor must keep a specified number of weapon systems operating. To the extent the contractor's repair system cannot maintain the required weapon system availability level, spare parts must be purchased. We assume a contractor is able to choose, to the best of its information, its repair capacity and other variables so as to maximize its profits (or utility, if the contractor is risk-averse) given the contract presented by the government.
We then developed a simulation of this model using parameters designed to be illustrative of real-world F-16 failure patterns and parts costs. We used this illustrative simulation to assess how a "maximizing" contractor might respond to various types of contracts.
One type of contract we examined was a type of per repair contract in which the contractor was directly compensated for repairs undertaken and spares required. The simulation results for this per repair contract were quite discouraging. A maximizing contractor has strong temptation to take advantage of this type of contract in a manner that is not in the best interest of the Air Force. For example, the contractor's incentive is to skimp on repair quality and costs and churn broken parts through its facilities, receiving a payment for every one. Our simulation suggests such a contract would prove to be very expensive for the government.
In the real world, the legal system, reputational effects, on-site monitoring of the contractor, serial-number tracking of parts, or other mechanisms can be used to avert this bad outcome. However, left unchecked, our model and simulation suggest per repair contracts have a tendency to induce low repair quality.
In light of this discouraging result, we sought to find contracting approaches that induced better contractor behavior.
Our results suggest the government's most cost-effective contracting approach closely resembles the C-21 and LN-15C availability-oriented contracts. Our simulation found the contract that minimized the government's expenditures, holding weapon system availability constant, was characterized by (1) a fixed payment to the contractor that did not vary with the number of units repaired; (2) a specified weapon system availability rate the contractor must meet; and (3) if the contractor is risk-averse, partial cost-sharing if expensive spare parts are needed to meet the required availability rate. This type of contract gives the contractor great incentive to do high-quality repairs; a reduction in the part failure rate does not result in a short-term reduction in contractor income.
This sort of contract also proves to be quite robust to informational asymmetries. An informational asymmetry, in this context, refers to a situation in which the contractor is better informed about aspects of the process than the government. Informational asymmetries seem plausible given that the contractor works day-to-day with the weapon system, while the government only writes repair contracts every so often. We obviously would not want a contract formulation that unraveled horribly if the government is moderately misinformed about the true characteristics of a situation.
For example, the preferred contract performs fairly well even if the government misestimates various parameters of the situation. To demonstrate this, we found the "true" best contract under various scenarios and compared its cost to the cost of the "mistaken" best contract, i.e., the contract the government believes to be, but is not, preferred.
|Government Informational Asymmetry||Government Cost Penalty %|
|Government believes failure rate is twice as large as it really is||0.1|
|Government believes failure rate is half as large as it really is||0.4|
|Government believes spares cost twice what they actually cost||710.9|
Table 2 presents the percentage cost increase above the cost-minimizing level if the government makes various errors. For example, suppose the government designed a "best" contract assuming the part failure rate was twice as large as contractors know it really is. Propitiously, assuming there is competition among contractors, the government's expected costs are estimated to be only 0.1 percent higher with this misunderstanding than they would have been if the government had optimized its contract for the lower failure rate. Similarly, underestimating the failure rate is not terribly costly to the government, beyond the fact that the government will need to pay the contractor a greater fixed payment to compensate for this increased failure rate.
Table 2's lone bad case is if the government grossly overestimates spares costs. It is worthwhile to subsidize expensive spares partially if contractors are risk-averse. The government does run considerable risk, though, if it so misestimates the costs of such spares that it is profitable for the contractor to rely on government-subsidized spares to maintain the required availability rate. One possible implication of this finding is that the government should be conservative in the expensive-item cost-sharing it agrees to, even with risk-averse contractors. Alternatively, the government can promise to pay a portion of the contractor's spares cost, but require the contractor to prove such costs.
In short, these results suggest that, even if the government overestimates or underestimates the costs of repair or the part failure rate, acceptable outcomes might still result, provided the contractor is well-informed and there is intercontractor competition. The only potentially dangerous informational asymmetry we found was if the government overestimated the cost of spare parts and offered the contractor overly generous spare-parts cost-sharing.
As noted, the appropriateness of this availability-oriented contracting approach depends on well-informed contractors who understand the repair costs and failure patterns of a weapon system. Hence, this sort of approach would seem better suited to mature rather than experimental systems.
Our model assumes there is competition between contractors for the contract to repair the system. Contracting is more challenging with proprietary technology. However, it is not unreasonable to think one could use an availability-oriented contract in such a setting. Litton, for instance, is the only contractor authorized to repair LN-15Cs.
More generally, modeling and simulation exercises of the sort we have undertaken are fundamentally an abstraction from reality. They provide suggestions and possible insights, but ultimate proof requires real-world implementation.
Fortunately, real-world examples, such as the C-21 and LN-15C contracts, are in place that fairly closely resemble the contracts recommended by our model and simulation. It might be reasonable to examine in greater depth the outcomes of these programs. If these programs have been successful, perhaps this sort of repair contract approach could be extended to other Air Force and DoD weapon systems.