Calculating the Costs and Benefits
The question of whether to proceed with SDI beyond the research phase into development and deployment depends not only on a determination that the various technologies are feasible but also on a calculus of costs and benefits. To attempt such a calculus, especially at this stage of the project, is to become aware of the difficulties in making plausible economic projections for complex technological projects with so many unknowns and variables.
The difficulties arise at every level of complexity. Even estimates based on the use of "off the shelf" technologies show wide variations. As we noted in chapter 1, advocates of the High Frontier scheme claimed in 1982 that its proposed space-based kinetic interceptor scheme would cost between $15 billion and $30 billion to deploy, but DOD analysts thought the same system would end up costing between $200 billion and $300 billion. Advocates of the three-layer Marshall Institute scheme estimate costs of development and deployment at $121 billion, but critics consider this estimate overly optimistic. Experienced analysts have produced projections of the cost of a full-fledged, multilayered comprehensive system that vary between almost $700 billion and $1 trillion.
Although cost projections of certain of the near-term technologies, especially those that would be ground-based, are apt to be less disparate than those for technologies still in a conceptual stage, all the calculations are based on assumptions that remain to be put to a practical test. The economics of space-based systems are the hardest to determine because there is simply no adequate basis for making firm estimates of the
cost of deploying and operating advanced sensors, weapons, and data processors in space—especially when the calculations involve "life cycle" costs, including launching, maintenance, replacement, and periodic renovation. Many of the most critical components are still undergoing testing and refinement in the research phase, which precedes development or engineering. Until the physical specifications can be spelled out in an "RFP"—the standard "request for proposals" by which the government invites bids from contractors—costs cannot even be approximated. Even then, estimates are apt to be uncertain, because unanticipated problems often arise in the development phase that raise cost estimates and delay completion. Military development projects typically end up costing substantially more than originally estimated—sometimes two or three times as much—and usually take longer than originally planned. SDI costs will also be affected by the expense of launching satellites into orbit, which could become lower as technology advances.
In view of these uncertainties, about all that can be done with any assurance now is to compare the original research strategy with the one that has actually been pursued. The more speculative prospects must also be examined in order to identify the issues that will need to be resolved before a commitment is made to development and deployment. Among the critical issues to be addressed, if development and deployment become feasible, are these:
Definitive answers to these and other related economic questions are inherently out of reach, and even educated guesses are doubtful while so many parameters cannot be quantified. At this stage, it is realistic only to raise the questions and consider some of the preliminary answers.
The Research Phase: Projections And Actual Appropriations
When the Fletcher panel was asked to prepare a detailed research agenda, it was also asked to provide two sets of cost estimates—one for a project that would be "technology limited" (or, in other words, that would allow the pace of research alone to determine allocations), the other for a "resource constrained" approach. Accordingly, the panel proposed a "technically aggressive" budget calling for the expenditure of $26 billion over five years beginning in FY1985, and an alternative, fiscally limited, budget of $14.3 billion. The panel endorsed the higher figure, which was announced publicly and incorporated into the DOD's five-year defense budget.
Had this projection been followed, the annual budget for SDI would have reached $8 billion by FY1989, approximately twice as much as Congress actually allocated. Over the five-year period actual appropriations for SDI, including those for work administered by the Department of Energy, have totaled $16.2 billion, or almost 40 percent less than the announced projection, and somewhat more than the lower, unannounced projection (see Table 3).
Although SDI has already cost as much in inflation-adjusted dollars as the Manhattan Project, it is not yet the "big ticket" item its advocates hoped, and its opponents feared, it would become. Still a research project, SDI does not yet carry as great a financial commitment as other military projects, which have passed into development and procurement. The Stealth bomber project, for example, could cost more than $60 billion if it is carried to full completion as presently planned. As a component of the Pentagon's budget for research, development, testing and evaluation (RDT&E), SDI currently accounts for 10 percent of the total—a large fraction, but one that was anticipated and is not so high as to deprive other high-priority projects of necessary support. The DOD had projected that spending for SDI would rise from 3.69 percent of defense R&D in 1984 to 13.06 percent in 1989. If the SDI budget had achieved that rate of increase, and if it were to continue to rise to an even higher proportion of total R&D expenditures, it would be taking
the lion's share of new resources for advanced projects—about 45 percent of cumulative increases projected for FY1987–1991.
Given the reluctance of Congress to allow the SDI budget to increase as planned, the immediate future is cloudy. An in-house concept paper on the strategic defense system issued August 1, 1987, ignored the political controversy and projected total research costs through a Milestone II decision on development of Phase I—presumably now to be reached in the mid-1990s—at about $50 billion. This presupposes a very sharp increase in the level of funding, one that would make the SDI research effort much larger than other DOD programs. As a Senate staff report pointed out: "If SDI proceeds on schedule by FY1992 the annual SDI budget request would be $11.5 billion or more than the Army or Navy now spend[s] on RDT&E for all their programs."
The Present Status Of SDI
According to its original rationale, SDI was supposed to be designed to determine the feasibility of moving to the next stage of development. But because of the crosscutting political pressures (described in chapter 7), the project has acquired a more ambiguous character. The timetable originally called for a decision to be made on development in the early to mid-1990s. In 1987 there were indications that the timetable had been moved up to enable a decision on development and simultaneous initial deployment to be made in 1991. The original plan also called for a full exploration of technical options so that the system to be considered for deployment would be one that would meet the announced goal of a comprehensive shield for population and not only for military targets. This goal has been significantly altered, under various pressures, to emphasize options for near-term deployment of a system that would be only partially effective (estimates range from 11 to 30 percent) in intercepting Soviet warheads. Because the proposed near-term system involves boost-phase interception by space-based kinetic kill vehicles (the space-based interceptor, or SBI) and by terminal ABMs with large footprints, it can be said to serve Reagan's announced goal of providing population defense, based on the "random subtractive" principle—i.e., the enemy RVs intercepted might be aimed either at military or civilian targets. Because this limited deployment cannot provide more than partial interception, however, it can hardly guarantee the safety of the population, much less promise to render nuclear missiles "impotent and obsolete."
Changes and uncertainties of this sort make it particularly hard to discuss the economic implications of SDI. Lieutenant General Abrahamson has been reluctant, for various reasons, to make cost estimates of future systems. The SDIO designed a "strawman" estimate based on a possible Phase I system. Its general view of what this option would cost is thought to be close to that of the pro-SDI George C. Marshall Institute, whose plan calls for a larger system that, it is claimed, would be more than 90 percent effective against the present Soviet missile force, at a cost of $121 billion—not counting the cost of air defense. The first SDIO estimate was more modest, calling for expenditures of $40–$50 billion, but this amount would buy only a very limited "first installment."
The most direct reason SDI has not met its projections is that it has encountered stiff resistance in Congress. But this opposition reflects a
skepticism that is widespread within the defense establishment itself, counterbalanced only by the self-interest of some defense contractors. Most technical specialists consider the project either not feasible at all or at too early a stage to warrant a commitment to development, let alone deployment. Military leaders are apprehensive that great increases in the SDI budget might deprive other strategic and conventional programs of the research support they need for improvements that are far more likely to be attained in the near term. Defense contractors are more supportive not because they are more optimistic about the technical prospects but because they fear that funding for research on strategic offensive weapons is likely to decline. During the Carter administration support for R&D on strategic offenses was running at about $2 billion a year. It has risen under the Reagan administration to about $8 billion—most of which has supported development of the B-1 and Stealth aircraft. If, as expected, that part of the budget should decline during the next few years, SDI could take up the slack and would therefore be particularly welcome to those dependent on military R&D contracting. Military industry clearly welcomes SDI, but mostly for its longer-range potential as a basis for future procurements. Procurements for strategic offensive forces rose to almost $20 billion in 1985 but are expected to decline to less than $5 billion annually by the late 1990s. Whether SDI's procurement potential will be realized depends on political decisions yet to be reached. The long-term implications of SDI thus remain highly speculative. Although the program has acquired some momentum because of the number and size of the research contracts, SDI's march into development and deployment is not a foregone conclusion.
Reorienting The "Investment Strategy"
The SDIO breaks down its expenditures in several different ways. One is by distinguishing basic research, exploratory development, advanced and engineering development, management support (including funds for installations like test ranges), and operational-systems development (or engineering development for systems already deployed). It is typical for DOD programs that only a small fraction—2.5 percent in the FY1985 request—is committed to basic research. All the research done by SDIO is classified by the DOD as "advanced development" for its purposes. Another breakdown, by type of work, divides the support among programs to develop kinetic-energy weapons (KEWs), directed-energy
weapons (DEWs), "SATKA" (surveillance, acquisition, tracking, and kill assessment), and battle management. The separate DOE budget supports research on the X-ray laser as well as on the SP-100 program—an effort to develop a 100 kw nuclear reactor suitable for space applications. In addition to the DOE budget, SDI also benefits from funds separately allocated to NASA for efforts to reduce launch costs.
The SDIO has defined its "investment strategy" as having three major directions. The first is to bring the most mature technologies to a point at which the implementation of a decision to deploy "would be largely one of engineering." The second is to pursue "emerging technologies" with potential for major improvements. The third is to ensure that investment is also made in "innovative ideas." The overall goal is to demonstrate technical feasibility and provide the broadest possible range of options. The office warned in 1986 that continued shortfalls in meeting its budget requests would have a substantial detrimental effect on the program. Further reductions would place SDIO in a position "where simply scaling back alternatives is no longer viable." Already, some programs have had to be canceled, others deferred and scaled down. In FY1987 a choice had to be made between the boost surveillance and tracking system and the satellite surveillance and tracking system. Because the first was designed for boost phase and the second for midcourse, the agency decided to support the former. Thus, what was conceived to be a more pressing need for a deployable first-phase system was chosen over a project that would be vital to a more comprehensive system. More generally, the shifts favor programs necessary for near-term deployment at the expense of those designed to enhance the "technology base."
In 1987 SDIO's budget was altered to reflect congressional budget cuts and to support increased emphasis on initial or early deployment. SDIO acknowledged in its budgetary analysis that its projected milestones reflect a change of relative emphasis away from countering the responsive threat and to support early-deployment options. The budget for near-term options was cut less than that for long-term options. In 1987 the request for KEWs was 20.8 percent, compared with the previous year's 20.5 percent. DEW meanwhile dropped from 33.6 percent of the FY1986 request to 26 percent of the 1987 request, and to 21.2 percent of the FY1988 request.
In this process of reorientation, the original criteria have been loosened. In defining its idea of the requirements of a defensive system, SDIO has not simply endorsed the Nitze criteria but has instead "reinterpreted"
them. Nitze laid down only two: survivability and cost-effectiveness at the margin. SDIO has made another consideration, "military effectiveness," the first criterion, contending that a defensive system must be able to destroy enough of an attacking force to make any adversary reluctant to try a preemptive strike. This objective would include the ability to deny the aggressor the opportunity of destroying a "militarily significant portion of the target bases he wishes to attack." The system must be capable of evolving over time to counter "responsive threats" or countermeasures and should be capable of deterring a strong offensive response. In other words, the aim is not to achieve population defense, at least at the outset, but to defend military assets so as to complicate an enemy's plans for a preemptive strike.
The second requirement is that the defensive system must possess "adequate [N.B.] survivability." This is said to mean that the defenses must maintain enough effectiveness to fulfill their mission even in the face of determined attacks and the loss of some components and not therefore present an appealing target for defense suppression. The offense must be forced to pay a penalty for trying to negate the defense. "This penalty would be sufficiently high in cost and/or uncertainty … that such an attack would not be contemplated seriously." In other words, the system need not be altogether survivable but only sufficiently so to enable it to perform its mission under attack.
The third requirement is that the defensive options generated "discourage" an adversary from overwhelming them with additional offense. The aim is to achieve defensive systems that could be maintained, protected, and proliferated more easily than an adversary could install reliable countermeasures: "This criterion is couched in terms of cost-effectiveness at the margin; however, it is much more than an economic concept." In other words, this restatement does not meet the Nitze criterion squarely but relies on a rough approximation emphasizing noneconomic considerations.
Contrary to this reformulation, cost-effectiveness at the margin is usually understood to mean that it must be less expensive (not just technically easier) for the defense to proliferate added defenses or otherwise add protection than it is for the offense to improve its threat whether by developing countermeasures or by adding more launchers and warheads. In the ideal case, a defensive system would have such a high kill ratio in each of its layers that the offense would have to proliferate missiles at an unacceptable rate to achieve sufficient penetration. Thus, a battle station should not cost more than the cost of the number of boosters
it can destroy. If a battle station could account for thirty boosters, as an OTA study suggests, "a cost of $50 million per booster would mean that the defense could spend $1.5 billion per battle station, and still keep up with the offense in the cost-exchange race." The trouble in applying this approach is that no such formula can be used to compute the actual future costs of development, deployment, and maintenance, because there is no comparable experience from which to derive cost estimates. In the case of offensive missiles, there is solid experience to serve as a guide. In the case of the defensive systems, it is necessary to make simplifying assumptions and estimate likely costs by extrapolating from an experiential base that is not fully comparable. As another OTA report aptly observes: "Nobody knows how to calculate, let alone demonstrate to the Soviets, the cost-exchange ratio between offense and defense."
Another problem is that what may prove to be an acceptable cost for one society may not be for another. Even though defenses may not be cost-effective at the margin in the narrow economic sense, the United States might be willing to pay for them on the theory that it could afford to spend more to defend itself than the Soviets can to maintain their threat. At the same time, because the Soviet Union is a more centralized society and demonstrably eager to maintain parity with the United States, it may be able and willing to spend more to improve its offense than U.S. taxpayers may be willing to commit to defenses.
Gauging The Direct And Indirect Costs Of Deployment
Ashton B. Carter has identified three ways of comparing the relative costs of offense and defense:
The Council on Economic Priorities (CEP) adds a fourth: a net cost assessment that includes complications caused by a Soviet deployment of defenses. (Thus, for example, a Soviet defense of strategic forces and
C3 could require a costly U.S. offensive response that must be added to the cost of a U.S. defense, insofar as the United States believes it advisable to maintain an offensive threat to those elements of the Soviet military forces.)
Although these formulas are useful, they do not preclude significant disagreements over cost estimates. Carter contends that fuel, in the form of chemical reactants necessary to pump chemical battle lasers, would require a full shuttle load to be orbited—a cost that, in comparison to Soviet cost per warhead intercepted, would not be favorable to the defense. SDIO analysts dispute this estimate, which they believe overstates the required weapon ranges and the power needed to fire a lethal beam at each target. Launch costs are potentially critical in view of the assumed weight of the battle stations to be orbited. In estimating costs, much would depend on the characteristics of the systems. If the Soviets elected to deploy missiles with less accuracy in order to threaten cities, costs would go down.
As Carter also points out, the offense can "stress" the defense by putting all its missiles in one region and increasing their number. Because of the need to provide continuous coverage, the defense is compelled to greatly increase—by as much as two or three times—the number of satellites needed to assure a successful rate of interception in the boost phase. Everything depends on assumptions made about the number of intercepts the battle satellites could perform in a given time and the time available during boost and post-boost phases. Countermeasures could also stress the defense, for example, by decreasing its effectiveness and increasing the weight that would have to be put in orbit.
Is SDI affordable? As the CEP study points out, in an economy producing more than $3 trillion worth of goods and services each year, a $32 billion six-year research program "is not likely to bankrupt our nation." But there is a significant opportunity cost to be paid, as the study also notes. Money invested in SDI is money that could be invested elsewhere, or, if unspent, would reduce the budget deficit. The costs could be significant for the balance of the economy, especially because of the possible effects on the distribution of technical employment in military and civil industries. Although the supply of engineers is affected by rising and falling demand in other defense and nondefense sectors, the supply of physicists is relatively fixed in the medium term. Efforts to draw physicists into work on directed-energy weapons, for example, would create shortages and reduce the effectiveness of R & D efforts elsewhere. Some of the slack might be taken up by encouraging immigration,
as higher U.S. salaries and better research opportunities attract qualified scientists from other countries. In general, the recession of the early 1980s cooled demand for qualified scientists and engineers, but the renewed growth in the economy took up this slack. Higher salaries in industry have already affected Ph.D. programs in the natural sciences. In 1971, 3,498 new Ph.D.s were awarded in the natural sciences. In 1981 the number was 2,528. An increasing percentage of Ph.D.s is being earned by foreign students (51.5 percent by 1981 compared to 29.8 percent in 1971). In the fall of 1980 universities reported that 10 percent of their faculty positions were vacant. While undergraduate enrollments in engineering increased by 66 percent from 1969–70 to 1981–82, faculty increased just 11 percent. Teaching loads were increased and schools were unable to offer courses in some areas. More of the actual teaching load has fallen on teaching assistants and part-time faculty.
The deep recession that began in 1981 mitigated the problem. Companies found it easier to recruit qualified personnel, and starting salaries fell back. In examining the cost of SDI deployment, the question for the future is what will happen if economic growth persists and accelerates and funding for SDI draws qualified researchers away from other employment. A study prepared for the NSF indicated that between 1982 and 1987, new entrants and immigrants would match or exceed the demand created by economic growth and attrition in all but three of the twenty-one occupational groups of engineers and scientists: computer specialists, aeronautical/astronautical engineers, and electrical/electronic engineers. If demand is high in both the civil and military economy, demand for computer specialists would exceed supply by 30 percent—or 138,200 people. But if demand is low, the problem would be much less serious. The most likely difficulties will be found in certain sectors, such as aerospace, where demand in both military and civil programs is expected to be high.
To gauge how many engineers, scientists, and technicians the SDI would absorb, if funding met original projections, the CEP team constructed a rough mathematical estimate, based on assumed yearly outlays for SDI, which were converted into production per industry. Then the average number of jobs per billion dollars of RDT & E expenditure was computed for each industry, on the assumption that SDI will require the same levels of production as the average in military RDT & E, which the CEP study admits is a questionable assumption, in view of the fact that SDI involves technological exploration rather than full-scale
development. The result is likely to underestimate the number of scientists, engineers, and technologists (SETs) required. The CEP study estimates that in 1984 the SDI required about 4,800 SETs, a figure that would have risen to 18,400 if projected spending over the next five years had materialized. Although SDI would remain only a small fraction of overall national and DOD demand, SDI could have used about 5 percent of the total SET personnel employed by the Pentagon and roughly 0.5 percent of the national total. The CEP estimated that between 1984 and 1987 the DOD would require roughly a third of all new engineers, of whom SDI would account for about 12 percent. SDI's impact would be particularly acute in the areas in which its work is concentrated: computer professionals and aeronautical/astronautical engineers.
By comparison, as the study points out, the Apollo program took up 13.5 percent of national funding for R & D, and NASA as a whole accounted for about 21 percent of total national R & D. SDI could take up 5.1 percent of total R & D by 1990. In marginal terms, however, SDI is already as significant as Apollo. Apollo received 40 percent of new R & D funding in 1965 and 20.8 percent in 1966. If SDI had met its projections, it might have used roughly 34.6 percent of new R & D funds during 1986. Even though the federal government encouraged university students to become space scientists and layoffs owing to defense cuts provided a pool of qualified people, by 1966 there were indications, nevertheless, of a severe shortage of scientists and engineers; the unemployment rate for engineers fell to 0.7 percent. As a result, R & D labor costs grew rapidly, and basic industries, forced to compete with aerospace and electronics industry salaries, found R & D investments less attractive. Similarly, increased demand for skilled manpower in the SDI could draw away skilled personnel from other fields, imposing higher costs on industries also requiring those specialties. The CEP study estimated that at the projected level the SDI could require 4 percent of all new engineers between 1984 and 1987 and that the DOD would likely take up a third of all engineers, with SDI requiring more than 12 percent of the total.
Projecting The Costs Of Initial Deployment
The projected cost of Phase I deployment has been a subject of some confusion. On March 19, 1987, Abrahamson testified to a Senate appropriations defense subcommittee that this deployment would cost between
$50 and $60 billion in 1985 dollars. In a September 1987 interview, however, he gave a considerably higher estimate, noting that it would cost at least $70 billion and perhaps well over $100 billion for an "initial, partially capable but very impressive deployment"—presumably one that was more capable and impressive than the initial deployment he was referring to in his congressional testimony. The second estimate was reportedly based on analyses prepared by SDIO for the Defense Acquisition Board in its Milestone I review. In February 1988 Abrahamson addressed the same issue and this time said that the cost would be between $75 and $150 billion (in 1988 dollars)—a figure that evidently included demonstration, validation, full-scale engineering development, and production. In the summer of 1988, he reported that technical progress would enable SDIO to construct a somewhat scaled-down version at a cost of $69 billion.
In reporting these changes, the Senate staff report of June 1988 noted that briefings given to the staff and confirmed by SDIO documents showed that the SDIO had come up with the higher estimate at least as early as June 1987. The report also points out that the panel reviewing Milestone I was aware of these estimates, pointing out that "by the time the necessary system and underlying technology work is complete, the design may change considerably and costs [may] change as well." As the staff report notes, in addition to these development costs, SDI would simultaneously incur other costs in support of ongoing research, estimated for FY1988–92 at more than $28 billion. Operation and maintenance costs for the Phase I system are estimated to be from $2 billion to $4.2 billion per year, and the Advanced Launch System to be used by SDIO would be developed jointly with the air force, which would contribute more than $1 billion dollars from other funds for this project. The budgetary impact would be sustained, as the report points out, because Phase II deployment is scheduled to begin "right on the heels of Phase I deployment." Directed-energy weapons would be deployed only two to four years after an initial deployment of kinetic weapons. The result would be "a mind-numbing annual expense," which, counting life-cycle expenses for both Phase I and Phase II, "could approach three-quarters of a trillion dollars."
These projections have been all but ignored in the political effort to build momentum behind SDI. The Reagan administration sought to keep the project on schedule by making the cuts required by the Gramm-Rudman-Hollings deficit-reduction law. This suggests an effort to keep SDI from economic comparisons, on the grounds that the preliminary
research is critical and must be undertaken within a short time frame, but it can also be interpreted as an effort to establish a momentum behind the program that will make it hard to terminate. Can Congress simply write finis to a program on which upwards of $30 billion will have been spent for exploration, or will it be tempted to provide at least some funds for development and deployment, even though the research does not result in a system that promises a virtually leakproof defense? Indeed, the pressure for early deployment of a limited or intermediate system suggests that this concern for tangible results is already being felt well before the research phase was supposed to terminate and make possible a decision regarding development and deployment. Now, the question is whether to proceed with deployment while the research phase continues.
Blechman And Utgoff's Four "Notional" Systems
Barry M. Blechman and Victor A. Utgoff suppose that cost estimates need to be made for four possible strategic-defense architectures. These are, first, a system aimed at providing terminal defenses, assisted by airborne surveillance and interceptor aircraft and designed to make U.S. nuclear retaliatory forces less attractive targets; second, adding a light, terminal area defense to the first system; third, protecting the entire population of the United States and Canada from ballistic missiles and aircraft by adding a space-based KEW component to the ground-based and airborne components used for the first and second systems; and fourth, addition of a space-based DEW component designed to destroy Soviet missiles in the boost phase with chemical lasers. These are designated, respectively, Alpha, Beta, Gamma, and Delta.
There are problems with this choice of options. The Delta system uses satellite-based chemical lasers, which have come into disfavor on various grounds. There is also no attempt to estimate costs for ground-based free-electron lasers, using space-based mirrors—an option considered more attractive than space-based lasers in critical respects—nor is there any consideration of the possible role in midcourse of neutral particle-beam discriminators. The study assumes that an initial, least expensive deployment would omit the boost-phase and aim for defense of military targets, whereas some actual proposals for an intermediate deployment would include boost-phase defenses and more than defense of military targets alone. Nevertheless, the exercise is a useful one. The
notional systems enable the analyst to make reasonably realistic estimates of the costs of putting satellites in orbit and maintaining them, hardening missile silos and satellites, and of the corresponding opportunity costs.
In developing their estimates, Blechman and Utgoff make certain critical assumptions. One is that interceptor missiles can be built—both the ground-based missiles and the space-based interceptors—that would have a probability of .9 of destroying their targets in any single engagement. If this is not possible, costs would rise considerably. A second assumption is that relatively inexpensive measures would be sufficient to protect the space-based components by hardening and orbiting as many as five decoy satellites for every space-based satellite deployed. A determined effort by the Soviets to develop effective ASATs could require more costly measures for defending satellites (including proliferation). Another assumption is that key components of all the systems would be built in sufficient numbers and according to the same design so as to capture the learning effects of manufacturing. They assume a 90 percent learning curve, or, in other words, that the marginal costs would decrease by 10 percent every time the quantity to be procured (e.g., of missiles) was doubled. If the assumed learning curve is only slightly higher or lower, the results would vary considerably. (For example, an 85 percent learning curve in one case would reduce costs by 25 percent of what they would otherwise be.) They also assume that during the fifteen or more years before a U.S. system could become operational, the U.S.S.R. could replace its force of offensive missiles and tailor its characteristics to make it less vulnerable to the defenses selected by the United States. Thus, the Soviets would deploy their missiles in an area roughly one-third the present size and would reduce boost times and harden their missiles against DEW.
The cost estimates of these four notional systems range from $160 billion for the Alpha to $170 billion for Beta, $770 billion for Gamma ($600 billion added to the cost of Beta), and $670 billion for Delta. The most expensive of the systems, Gamma, would therefore cost on the order of $44 billion annually during its ten peak years. Although this would represent a 15 percent real increase in the current level of defense outlays and a commitment of roughly 1 percent of the nation's resources for a considerable period of time, defense expenditures would increase from their current level of about 6 percent to only 7 percent. As they point out, this figure has been far exceeded in wartime and been matched or exceeded during all but a few of the peacetime years from 1945 to
1970. Only the first two systems could be financed within historical levels of spending for strategic offensive forces; the space-based systems would require greater levels than have been historically committed for strategic forces. Trade-offs will also have to be made, no matter which level of expenditure is chosen, with other defense requirements. The $160 billion required for the Alpha system is equivalent to the cost of eight aircraft-carrier battle groups or twenty-seven wings of F-15 fighters or fourteen armored divisions. The more expensive systems would cost far more than the army now spends on development and procurement and would be roughly comparable to the air force budget for development and procurement.
If strategic defenses were to be financed without harming other defense programs, cutting civilian programs might be an alternative, but the cuts would have to be significant. Peak expenditures for the Alpha system would be roughly the same as current federal outlays for higher education. The Delta and Gamma costs of $44 billion and $37 billion per year, respectively, are comparable to the $25 billion devoted to farm price stabilization in 1986, the $30 billion for health care services, and the $71 billion for Medicare. Funding the Delta system in its peak years would require cutting about 20 percent of the current $180 billion in so-called discretionary nondefense federal spending.
Another possible alternative for financing strategic defenses would be to increase federal revenues. This would require, for the Gamma system, roughly an 11 percent increase in federal revenues from individual income taxes, based on 1985 returns, or about $570 for the average family earning between $30,000 and $50,000 per year. Alternatively, the system could be financed by raising corporate income taxes by 50 percent. Either step would cut individual consumption and saving and affect economic growth, employment, and inflation.
Blechman and Utgoff note, however, that increasing the share of federal expenditures allocated to strategic defense would not necessarily have adverse economic effects. As they point out, history suggests that the effects of defense expenditures can be influenced by monetary policy and by international economic circumstances in such a way as to counterbalance negative effects. They agree with the CEP group, however, that specific industrial and scientific sectors could be affected by the manpower and resource requirements of SDI. A large program "could strongly impact on the availability and price of certain kinds of scientists and engineers, computer programmers, and other specialists" and distort markets for specific types of raw materials and manufactured
goods. Such adverse consequences could be minimized by cooperation between the federal government and affected industries and federal encouragement of related occupations.
The notional Alpha system—a terminal defense—would be based on known technology and could be developed within seven years and be fully deployed in roughly double that time. It would render the U.S. deterrent force more resistant to attack by raising the cost of attacking the facilities where the forces are based—more nuclear weapons would be required to destroy each base than could be expected to be found at the base at the time the attack was executed. Potential attackers would thus lose the incentive to attack because they would use up more warheads than they could reasonably expect to destroy. The system would employ ground-based interceptors (HEDI and LEDI) to provide two layers of defense and would be assisted by ground-based radars and airborne systems for surveillance, target acquisition, and battle management. Early warning would be provided by surveillance satellites in geosynchronous orbits and ground-based over-the-horizon radars. The surveillance systems would require no incremental expenditures.
The missile-defense system would be accompanied by an airborne system to defend against bombers and cruise missiles. Surveillance aircraft would continuously patrol a barrier well to the north and off the coasts of the United States. Shorter-range interceptor aircraft guided by AWACs-type aircraft would intercept enemy bombers. (Airborne air defenses were chosen because they would be less costly than ground-based systems and more flexible in that they could intercept at the points of attack, whereas SAMs would have to be deployed in sufficient numbers all along potential attack corridors.) Because the terminal-defense system is designed to function within the atmosphere, the decoy problem is ignored, and it is assumed that any warheads that escape interception by the high-endoatmospheric system would be destroyed by high-acceleration, low-endoatmospheric interceptors before they could explode close enough to military targets to destroy the missiles, bombers, and submarines at which they are aimed.
The sizing of the air defense assumes that the U.S.S.R. deploys a force of two hundred Bear F and Blackjack strategic bombers carrying an average of ten cruise missiles each. Estimating that two-thirds of the bombers would participate in the initial attack, Blechman and Utgoff add that by the turn of the century the Soviets might also be able to launch 350 cruise missiles from submarines off the U.S. coasts. The United States' defense would require three hundred shorter-range interceptors
and ten AWACs already planned for continental air defenses in the 1990s. New air bases would be required, and these bases would also have to be defended against both air attack and ICBM attack.
According to their estimates, the total number of U.S. weapons considered vulnerable to air attack would be 2,943, assuming that silo-based missiles will have been launched before cruise missiles launched by bombers could hit. Because the Soviets would have available 1,700 bomber and submarine-launched cruise missiles, the defender could expect to kill 600 of the attacking cruise missiles by interceptors, assuming that 300 U.S. aircraft were available each carrying four missiles and that the alert rate of the aircraft and the kill probability of each missile were both .7 (300 × .7 = 210; 210 × 4 = 840; 840 × .7 = 588). The remaining 1,100 would have to be destroyed by aircraft in the outer barrier. If one calculates expected kill ratios, sixteen alert surveillance aircraft would be required, in addition to the already-planned air defense forces. To allow coverage of the entire 4,500 nautical-mile-long perimeter, a total force of fifty aircraft would be required (assuming a 40 percent strip alert rate). Because of the submarine threat, however, additional planes would be needed, raising the inventory requirement to seventy-five armed surveillance aircraft. The effectiveness of the aircraft would be greatly improved by aerial refueling; tanker aircraft would also be needed, totaling thirty-five, the equivalent of the KC-10A. The air bases for this defending force would have to be dispersed and as well defended as the missile silos. Thus, thirty-two bases might be advisable for the 275 strategic-defense aircraft of the Alpha system, in addition to the ten already available. Each base would need to be defended by 122 HEDI and 20 LEDI and 1 ground-based radar. The targets would have to be hardened against salvage-fused missiles lest nuclear effects destroy targets such as hangars and aircraft.
Of the total cost of $160 billion, somewhat more than half would go for air defense. If the aim were to defend only silo-based ballistic missiles, and to defend them only from attacks by ballistic missiles, the total ten-year cost for a terminal system would be roughly $30 billion. This system could be less expensive if only ground-based radars and a single layer of defense were employed.
The SDIO/Marshall Institute Phase I System
By comparison, the strawman concept advanced by the SDIO in 1987 contemplated a system using three thousand space-based kinetic-kill
vehicles carried on several hundred satellites—and both ERIS and HEDI terminal defenses supplemented by airborne tracking systems but not by an air defense. Such a system—which was not put forward as just a defense of military targets but, instead, in a random subtractive mode—would cost between $40 billion and $60 billion, according to SDIO estimates, and could be deployed as early as 1994 and 1995. The system was estimated to have an effectiveness of 11 percent against the current Soviet ballistic missile threat. A representative for Lockheed Missiles and Space claimed that the company's ERIS missiles could be deployed within a few years and could provide effective interim defense against the current Soviet threat at a life-cycle cost of $1 million per intercept.
The Marshall Institute proposal, which is thought to be closely modeled after the one SDIO is contemplating (and therefore warrants more scrutiny than it might otherwise merit), calls for a more extensive deployment than the SDIO strawman proposal. Arguing against the common view that terminal defenses are the most feasible in the near term, followed by midcourse defenses, the proposal contends that technical progress achieved under SDI had "invalidated this traditional view." Instead, the proposal suggests that a boost-phase defense can be deployed in the same time frame as the ground-based layers.
The Marshall defense would be based on homing interceptors—non-nuclear, heat-seeking missiles using the same technology developed for air defense. The system would have three layers, boost phase, late mid-course and terminal interception. All would rely on kinetic weapons in view of the fact that beam weapons lag "a few years" behind KEWs. Deployment of the full defense could be achieved seven years after the decision to deploy, assuming "streamlined management and procurement procedures are followed." Deployment of the ERIS layer would begin five years from the date of the decision to deploy. A decision to proceed with an incremental deployment would provide the earliest possible protection against "accidental or irrational" launches and a useful degree of deterrence against limited attacks on key military sites. The institute cites SDIO's success in completing the Delta 180 experiment in fourteen months from initiation, claiming that normal procedures would have required twice as long.
The level of effectiveness of the three-layer defense is calculated to be approximately 93 percent against a "threat cloud" of 10,000 warheads and 100,000 decoys. With decoy discrimination in the ERIS layer, the proposed defense is presumed to have an effectiveness "well in excess of 99 percent." Defenses of this level of effectiveness "will virtually foreclose
the possibility of a nuclear first strike against the United States." The detailed calculations supporting this conclusion are based on the assumed present strength of the Soviet strategic force as well as the assumed kill probabilities of the proposed U.S. defense. The Soviets are believed to deploy eight thousand to ten thousand strategic ballistic missile warheads—a total that may grow to as many as fifteen thousand in the mid-1990s, according to CIA estimates. Currently, if the Soviets were to assign two warheads to each target in the United States, they could be confident of destroying it. With a U.S. defense in place, an attacker would have to devote more than two warheads to each target to achieve the same kill probability. The Marshall study calculates that if the defense is 90 percent effective, thirty-eight warheads would have to be assigned to each of the targets to assure the same level of destruction. The statistical calculation supporting this estimate is not provided, but presumably it is based on the assumption that because the attacker cannot know which of the warheads he launches will be intercepted, he must compensate by greatly increasing the size of the attack. Because the United States has more than one thousand high-priority military targets, this calculation leads to the conclusion that the Soviets would have to launch more than 38,000 warheads to achieve the same results that they could now expect in the absence of U.S. defenses. This would be more than three times the present Soviet inventory and "an impractically large number." Even so large a force, moreover, would not be capable of destroying the U.S. arsenal on submarines at sea or bombers on alert. Thus, the proposed defensive deployment would introduce "a paralyzing degree of uncertainty" into an adversary's strategic calculus, virtually ruling out consideration of a possible preemptive strike.
The cost of such a system to the point of initial operating capability (IOC) or partial deployment is estimated to be $54 billion; full operational capability (FOC) would cost $121 billion. Thus, the annual cost during development and deployment would be in the range of $10 to $15 billion annually, or between 3 and 4 percent of probable DOD annual budgets through the mid-1990s. The most expensive of the three layers would be the space-based one. The cost to manufacture each SBKKV is estimated at $1.5 million, based on independent studies by three contractors and the DOD. The cost to launch for each SBKKV is put at $.75 million, based on launch costs of 500 lbs. at $1,500 a lb.; the launch cost of the satellite that carries the SBKKV is also estimated as $.75 million. Cost of operations and maintenance for a ten-year life cycle is projected to be $1.5 million per SBKKV. The sensor satellites in
low earth orbit (LEO) are estimated to cost $1 billion each, for a total cost of $10 billion for the ten sensor satellites. Cost of four sensor satellites in geosynchronous orbit (GEO) is estimated to be $2 billion each, or $8 billion for all four. The space-based layer, comprising eleven thousand SBKKVs and supporting sensor satellites would cost $68 billion. The ERIS layer would cost a total of $32 billion for ten thousand interceptors, including the aircraft carrying the airborne optical system (costing $10 billion). Each ERIS interceptor would cost $1.5 million along with $.75 million for launch facilities and operation and maintenance costs for ten years of $.75 million. The HEDI layer would cost a total of $18 billion. Each HEDI would cost about twice as much as an ERIS missile, $3 million each, because a costlier rocket is required to produce the higher acceleration and to cool the interceptor detector window. IOC would include deployment of sensor satellites, AOS aircraft, ground radars, and one-quarter of the full complement of SBKKVs.
The institute calculates that the system would enjoy a favorable marginal cost ratio compared to Soviet ICBMs because their cost is approximately $20 million per warhead. In the boost phase a U.S. defense consisting of eleven thousand U.S. SBKKVs, estimated to cost about $68 billion, would in theory destroy boosters and buses carrying a total of nine thousand warheads, costing the Soviets $180 billion, giving the defense a marginal cost ratio of 3:1. For ERIS, the equivalent cost of each missile is approximately $3 million. But its effectiveness is estimated to be 70 percent, so its net cost would become $4.2 million. Cost of the incoming warheads is $20 million each. If ten decoys are deployed per warhead and decoy discrimination is not available, $42 million in ERIS missiles must be expended to destroy one warhead, with a marginal cost ratio of approximately 2:1 in favor of the offense. With effective decoy discrimination, the cost ratio shifts to 5:1 in favor of the defense. For HEDI, including ten-year life-cycle costs, each interceptor and its launchers are conservatively estimated to cost $6 million. At an average effectiveness of 70 percent, the cost would be $12 million per interceptor, or a cost ratio of 1.7:1 in favor of the defense. Against submarine-launched missiles on flat trajectories, HEDI effectiveness would rise to 100 percent and the cost ratio would be 2:5 in favor of the defense.
Former secretary of defense Harold Brown says that the proposal for a three-tier defense is "unrealistic and would not lead to an effective capability." The KKV architectures, he observes, appear to be too easily countered by a reactive offense. The 1993–94 initial deployment date, moreover, is "difficult to credit." Six years is the same time that elapsed
between the decision in 1981 to produce the B-1 bomber and its initial deployment. Seven years is about the time it will take to deploy an aircraft carrier if Congress appropriated funds for it. An SDI system, even in its initial form, is far more complex and unprecedented. Although the Soviets are not now developing a fast-burn booster, a U.S. decision to deploy a boost-phase defense would provide a strong motivation for such a development—along with space mines and other countermeasures. According to Brown, the countermeasures can be developed more cheaply and deployed more easily than fast-burn boosters.
In contrast to the Marshall Institute plan, the study by Blechman and Utgoff estimates that space-based interceptor missiles "could probably be fully deployed around the year 2012." These are the same SBKKVs the Marshall study sees as fully deployable a decade sooner. The study assumes that by the time the U.S. defense could be deployed, the Soviets would have altered the characteristics of their offensive arsenal so that their missiles would spend only 90 seconds in boost phase, that the post-boost phase would also be shortened to 60 seconds, and that Soviet missiles would be deployed in a more compressed area. On the U.S. side, the main difference in architecture between the Blechman and Utgoff study and the Marshall proposal is that the latter provides for no special battle-management satellites—eliminated presumably because the decentralization of the system would make them unnecessary. Blechman and Utgoff also assume that the interceptors would be capable of achieving a kill probability of 0.9 against any Soviet missile within range, but in their scheme the battle-management system would serve a critical function of assigning targets and ordering second strikes against targets not killed in the first wave.
The Blechman and Utgoff study also assumes a 5 kg homing warhead with an initial missile weight of 150 kg in order to achieve a burnout velocity of 10 km per second. The study estimates that 60,800 interceptors would be needed against the current force of Soviet ICBMs, 86,900 if intermediate-range missiles are also considered. The estimate calls for placing no more than 175 interceptor rockets on one satellite against only ICBMs—250 if intermediate weapons are included. Because the Soviets could cut a hole through the defense by destroying as few as nine to twelve of these battle satellites, it is probable that no more than fifty rocket interceptors would be placed on one pod. This also guarantees that enough interceptors would be close enough to the offensive missiles to enable a timely arrival of the second wave of interceptors. The system would also include decoy satellites—five for each battle satellite. In all, a
total of 1,335 battle satellites and 6,675 decoys would be required to defend against Soviet ICBMs and SLBMs; 1,915 satellites and 9,575 decoys against a total force, including intermediate missiles. The payload weight would be 7,500 kg, the total satellite weight that would have to be placed in orbit would be 15.3 million kg, or 21.9 million kg to deal with the entire Soviet missile force. Given the roughly 16,000 kg payload the current space shuttle could put in orbit, launching these weights would require approximately 960 and 1,370 shuttle-equivalent flights respectively. Repairs and "rolling renewal" would require annual launches of about 10 percent of the initial weight.
In contrast to the launch estimates of the Marshall Institute, Blechman and Utgoff contend (on the basis of a study by R. G. Finke and others) that the initial life requirement for this system would not be great enough to justify the high costs of developing and installing a new, large unmanned booster and would be better met by accepting the marginal costs of upgrading the shuttle. An upgraded shuttle would be capable of launching twice as much payload as the current shuttle: 60,000 kg in a low earth orbit required for the interceptors, 33,000 kg into the polar earth orbit required for the battle management satellites. Some 580–820 flights would be required over a four-year period, and approximately 50 or 70 per year thereafter. Based on the cost of producing ASATs now being manufactured by the LTV Corporation—taking into account reductions for increased numbers—the study estimates that the cost of producing the 73,600 interceptors needed (including 10 percent extra for testing and ground-based maintenance) would be $180 billion; $246 billion for the larger system. This represents an average cost of $2.5 million per missile. Research and development would cost $3 billion, including $2 billion for the battle satellites. The cost per satellite would be $85 million, with reduction of cost as the design was standardized. The cost of producing battle satellites would therefore be about $42–$50 billion for the entire system, and another $20 billion for twenty battle-management satellites. It would cost $33 billion to launch and sustain the space-based component, or $8 billion for the larger system. Launch costs thereafter would be $2.8 billion to $4 billion. Thus, the total cost for the space-based element would be between $460 and $600 billion (roughly 60 percent of the cost attributable to interceptor missiles). And it would cost an additional $170 billion for the ground-based system that would complement it. Given other marginal costs, the total system cost would be between $630 billion and $770 billion, or far more than the Marshall proposal contemplates.
The Spin-Off Issue
As to the possible spin-off benefits, the CEP study noted that it is difficult to ascertain the value of such spin-offs even from the space program, where the effects can be studied. It seems clear that NASA investments have paid off handsomely in promoting the commercial use of communications satellites. Other examples could also be cited, notably the use of satellites for mapping and exploration for minerals as well as weather forecasting. But the quantitative economic benefits are hard to measure, particularly because many of the benefits are captured by world users in general, rather than solely by the United States.
Most analysis seems to show that government-supported R&D is less likely to promote economic efficiency than industry-sponsored R&D. Government support may help develop products for particular uses, but these may not be commercially applicable. Some studies also suggest that firms cut back on private funding of research when they receive government research support. Economists also generally assume that basic research is likely to have more favorable spin-offs than applied research, whereas SDI emphasizes applied research. Private commercial applications of high-energy lasers, particle beams, large optics, and infrared sensors are "not immediately obvious," as the CEP study by William D. Hartung and others notes. Potential spin-offs are also limited by extremely tight security. SDI has tried to alleviate these problems, as NASA did, by setting up a special office to promote commercial spin-offs and by allocating funds, in the case of the X-ray laser, for research into medical applications, but it is not clear that SDIO can do much to resolve any problems of dissemination that might result from the concentration of this funding on military projects.
In its 1987 report to Congress, SDIO addressed this issue in discussing the work of its Office of Civil Applications. Emphasis was placed on the development of a referral data base containing synoptic data regarding new and unique SDI-generated technologies, which would be made available by computer modem to qualified business and academic clients approved by the DOD. In addition, a subcommittee is to be established for the SDIO Advisory Committee, with technology applications panels to function on a continuing basis "in reformatting technology into industrial technology profiles, identifying potential applications, reviewing client inquiries, and recommending further development of research." These panels are being established in several generic areas: biomedical applications, electronics, communications and computer applications;
power-generation, storage and transmission applications; and materials and industrial-process applications. In 1984 Congress included a program for medical applications of the free-electron laser in the SDI program. Five regional medical free-electron centers are being established: three at universities and two at nonprofit laboratories (Brookhaven and the National Bureau of Standards). Other laboratories are being involved in the general effort.
According to SDIO, a number of key areas for civil applications have been identified: computer data-processing speed and efficiency enhancements through improved components, circuitry, and software; lighter, smaller, more capable and energy-efficient electronic components; software with artificial intelligence that would allow computer systems to learn from experience and to make realistic deductions; optical computing using laser light; electrical power systems that are more efficient, and less expensive sensors that are lighter, smaller, more sensitive, and less expensive for medical applications, manufacturing, research, control systems, and many other applications; cryogenic cooling systems that are lighter, smaller, and more efficient for food preservation, medical applications, etc.; lightweight mirrors with computer-controlled adaptive alignment for laser applications to manufacturing processes; electrical systems hardening techniques applicable to reducing or eliminating noise and other interference in communication systems; tracking and pointing technology that might be useful in commercial aircraft guidance and control and ground-traffic monitoring; tomography-assisted technology to enhance medical diagnosis; free-electron applications for noninvasive surgery and diagnosis; integration of laser technology, robotics, and computerized precision control into industrial and biomedical applications.
Again, however, the opportunity costs need to be included in the balance. SDI could have a significant impact on R & D resources. The Apollo program consumed a much larger fraction of total U.S. R & D, especially in its peak year of 1966, than SDI will. NASA as a whole took up 20.8 percent of all research and development spending in the United States. In 1965 and 1966 Apollo required 40 and 20.8 percent respectively of new R & D resources. In 1990, even in the unlikely event that original projections are met, SDI would take up only 5.1 percent of all national R & D. Prior to the Apollo program, NASA did not significantly compete with private industry for scientists and engineers, nor did NASA's early growth strain the labor supply because of an economic slowdown in 1960–61. And NASA made a major effort to expand the
labor pool so as not to compete with the private sector by establishing cooperative training and recruiting relationships with major universities. By 1966, however, there were unmistakable signs of a severe shortage of scientists and engineers. This competition, according to the CEP study, "exacted a high toll on U.S. industry during the 1960s." The cost per employed scientist and engineer rose for civilian industry. Labor costs expanded as much as 80 percent for one industry, and almost as much for others. A greatly expanded SDI project could have similar results.
Still, if the NASA experience is repeated, SDI could also have significant economic benefits. NASA's R & D is thought to have been "strongly associated" with rising productivity in the economy between 1960 and 1974. NASA's R & D was said by a Chase econometrics study to have yielded a high rate of return in increased productivity; this study was performed for NASA. A GAO account of it, however, found that the claim was based on sensitive assumptions. The GAO did not claim to have disproved the study but only suggested that macroeconomic studies were hard to validate. Case studies are more persuasive, but they are inevitably too particularistic to allow for generalization. A study by the Midwest Research Institute in 1971 found a variety of innovations derived from NASA efforts. NASA's research provided the basic knowledge that made possible geostationary satellites now used for civil and military communication. NASA was the initial user of many industrial products. NASA assisted industry by evaluating alternative technologies, demonstrating new devices, and reducing uncertainty. "Nowhere," as the CEP notes, "is the value of NASA efforts more evident than in the development of communications satellites." Satellite communications is an application that would certainly not have been developed as rapidly or as well if not for NASA's pioneering efforts.
Such anecdotal evidence cannot prove that NASA's overall impact has been positive for the economy or that defense research is also of net benefit to the civil economy. Econometric analysis has shown, for example, "that while industry-financed R & D is significantly related to productivity, government-supported R & D is not statistically connected to increased economic efficiency." And industries that are major recipients of federal R & D funds have the lowest productivity return on R & D from both sources. The reasons are several. Government R & D may lower the marginal cost to the firm of performing R & D and thus expand the amount of R & D that the firm can perform. Most of the government's R & D dollars are spent in R & D-intensive industries. In
this way, government support could stimulate civil R & D. Second, most of the government's R & D goes for developing products for its own use, such as aircraft and communications equipment, not on improving the production, for example, of commercial foods. There is an important distinction between product and process improvements. Product improvements are useful only for the specific product; process improvements tend to be diffused throughout the economy. For this reason, government defense R & D is "a rather inefficient way to increase economic efficiency" because Pentagon products are not designed in a cost-minimizing environment. Another study has shown that increases in federal R & D were associated with reductions in company-financed research, contradicting the supposition that federal R & D stimulates greater private R & D. Instead, it tends to crowd out commercial research. A federally funded increase of one hundred R & D scientists and engineers one year will result in a decrease of company sponsorship of thirty-nine within a year, essentially no change the next, and an increase of seven the year afterward. Thus, the evidence is inconclusive.
Another issued raised by the CEP study is the contention that spin-offs result more from basic research than from applied research in defense R & D and that most of the work done under SDI will be applied. Most economists tend to believe that if the effort is aimed at producing new technologies for specific uses, fewer commercial spin-offs will result. All of the SDI programs are designed to produce specific items for the layered defense against ballistic missiles. If in the process progress is made in developing lasers, sensors, and computers, that progress will not necessarily be transferable to commercial products and processes. There are instances where parallel uses are apparent and significant, notably the development of very high speed integrated circuits (VHSIC), which are necessary for real-time data processing for strategic defenses and have already found commercial markets. But there is far less likelihood that research on an electromagnetic railgun will prove adaptable to civil transportation, for example. Another problem inhibiting technology transfer to civil products is security classification. This concern with security has already blocked the transfer of VHSIC technology to civil applications lest the Soviets use a prototype to "reverse engineer" a device incorporating it.
Because SDI has not received the level of funding called for by the Reagan administration, it has not had the impact on defense spending or on the civil economy that has aroused the keenest controversy. If a decision is taken to deploy even a modest BMD system involving space-based
defenses, the likely expenditures would be very high, whether one accepts the estimates of the advocates of such systems or of those who can claim to be more objective. It is evident that any full-scale space-based system would cost hundreds of billions of dollars to deploy, and billions more for a complementary air defense. Over a period of years even expenses in the range of hundreds of billions of dollars can be assimilated, but only if other military and civil programs are constrained. Such a program is not likely to have great positive benefits for the civil sector but is more likely to draw talent and other resources away from it and to raise factor prices in a strong economy. It could also generate a momentum in favor of continued expenditures, which would perpetuate dependency on defense spending in key industries and regions. Thus, even if deployment eventually becomes advisable for strategic and political reasons, there should be no illusions about the likely costs or about the potential spin-off benefits.