Pages 12-13

In 1960, while working for General Gerhart, Glenn scored one of his greatest coups. The Secretary of Defense at that time, Thomas Gates, was interested in finding a means for imposing more cohesion and integration on the nation’s operational plans for employing nuclear weapons. Threatening the Soviet Union with a retaliatory nuclear attack had for some time been the centerpiece of U.S. efforts to deter a Soviet attack on the United States. Despite this, U.S. plans for nuclear attacks on the Soviet Union and the Warsaw Pact were not being developed in an integrated manner. Rather, they were developed by the individual regional combatant commands and were not well coordinated.

Charged by General Gerhart to involve himself in this problem, Glenn played a pivotal role in conceiving of a process by which a joint planning staff would develop an integrated plan. He called it the SIOP. Working with Gen Thomas Power, who was then commander of SAC, and Col George Brown, who was military assistant to Secretary Gates, Glenn helped to choreograph the process for gaining the secretary’s approval of this concept (see “The Advent of the SIOP,” pp. 22–30). One result of the SIOP triumph was that General Power, who had a reputation as fearsome as LeMay’s, saw to it that Colonel Kent was eventually promoted to brigadier general, in 1963.

[…]

In the summer of 1962, Glenn was assigned back to the Pentagon. This time, he was working not for the Air Force but in the Office of the Secretary of Defense (OSD). Specifically, he worked for Dr. Harold Brown, who was then the Director for Defense Research and Engineering (DDR&E). This was the heyday of Robert S. McNamara and the “Whiz Kids”—analysts brought to OSD from places like RAND. They were recruited in an effort to bring greater rationality to defense decisionmaking through the application of systems analysis. In Dr. Brown, Kent had a boss with a superb grasp of mathematics and defense planning. Dr. Brown also enjoyed the trust and confidence of Secretary McNamara. The Office of Systems Analysis was headed by Charles Hitch and was home to many of the whiz kids. A competition of sorts emerged between DDR&E and Systems Analysis.

Glenn had impressed Brown with some early work he did on issues relating to the defense of the United States against nuclear attacks. In time, Brown directed Kent—then a brigadier general—to analyze the utility of the full range of programs for limiting damage to the United States from a nuclear attack—offensive weapons and delivery systems, active defenses (air and missile), warning systems, and passive defenses (such as fallout shelters). This effort consumed General Kent’s time for the better part of a year. The study’s primary conclusion was that the U.S.–Soviet strategic balance was dominated by offensive systems and that any investment the United States might make in systems designed to limit damage from a Soviet attack could be overwhelmed at less expense by a larger attack. This insight led to a shift in policy that emphasized ensuring a secure second-strike capability over limiting damage and laid the conceptual foundation for two-sided arms control.

Pages 22 – 30
The Advent of the SIOP

In the 1950s, I was a colonel heading the Weapons Plans Division on the Air Staff. This division dealt solely with nuclear weapons and related issues. It was subordinate to the Director of Plans, Maj Gen Glen Martin. He, in turn, worked for the Deputy for Plans and Programs of the Air Staff, at the time Lt Gen John Gerhart.

The United States introduced nuclear weapons into its operational plans (OPLANs) in an incremental and less-than-integrated manner. Different entities developed their own OPLANs without much real coordination with others. The Air Force developed a fleet of bombers capable of carrying nuclear weapons. The Navy built submarines that could launch intercontinental ballistic missiles (ICBMs) and also carrier-based aircraft that were nuclear capable. Later, the Air Force created its own array of ICBMs. The incremental nature of these developments resulted in a less-than-integrated plan for how these weapons were to be employed. The United States had no workable approach to unify the planning among the various commands involved.

One day, General Gerhart called me to his office. “There is an ongoing item before the Joint Chiefs of Staff [JCS],” he stated. “The Chief just told me that we cannot afford to lose on this matter. I am hereby making you the ‘action officer.’ Forget, for a while, about running your division. Focus your entire efforts on winning this action before the Joint Chiefs.” [*]

[* – For more detail on this debate, see Nathan Twining, Chairman of the Joint Chiefs of Staff, memorandum to the Secretary of Defense, “Target Coordination and Associated Problems,” August 17, 1959.]

The action before the JCS had been prompted by a memorandum from the Chief of Staff of the Air Force to the JCS recommending that, to have an integrated OPLAN for nuclear strikes on the Soviet Union, the Strategic Air Command (SAC) should assume control of the Navy’s Polaris submarines. The Navy and the Army were vigorously and emotionally opposed to this recommendation. SAC was seen (rightly) as an arm of the Air Force, and the Navy, in particular, had a tradition of jealously guarding its autonomy. The fact that the Air Force, the upstart among the services, had grown rapidly under the Eisenhower administration intensified the determination of the other services to prevent this expansion of SAC’s authority. [*]

[* – In 1959, the Air Force’s share of the DoD budget approached 50 percent. Since 1965, it has never exceeded 39 percent.]

The next day, I discussed the matter further with General Gerhart. I told him that the real issue was whether the United States should have one integrated OPLAN or should just coordinate three plans: one developed by SAC; one developed by the U.S. Navy Commander in Chief, Atlantic Command; and one developed by the North Atlantic Treaty Organization (NATO).

I then pointed out that, to have a single integrated plan, it was not necessary to place the Navy’s Polaris submarines under SAC or a unified command. Instead, I recommended that the Air Force advocate the formation of a joint strategic planning group. Personnel from all services would staff this group. The commander of SAC would head the group, and his deputy would be a Navy admiral. I also proposed that this group be collocated with SAC in Omaha, Nebraska.

If we want to prevail, I told General Gerhart, we must abandon the idea of placing Polaris submarines under the control of SAC or even some other specified command, a step that would provoke strong and emotional opposition from the other services. Instead, we should focus on the critical issue of a single integrated operational plan (SIOP). We could have a SIOP without grappling with the issue of command and control over nuclear forces. General Gerhart agreed with this approach and recommended to the Chief of Staff of the Air Force, Gen Thomas White, and the commander of SAC, Gen Thomas Power, that the Air Force change its position accordingly. They agreed, though General Power was reluctant, and so the Air Force came to advocate a SIOP and a Joint Strategic Targeting and Planning Staff (JSTPS) in Omaha. [*]

[* – Thomas White, Air Force Chief of Staff, memorandum to Secretary of Defense Thomas Gates, with attachment on Strategic Targeting Authority, June 10, 1960.]

Both the other services opposed our new proposal. In part, this was because the previous attempt to gain operational control had poisoned the well. Even without this, however, the other services would have been suspicious of the proposal because they saw the SIOP and the joint planning staff in Omaha as an ill-disguised gambit by the Air Force to dominate the planning and conduct of nuclear operations, and they were determined to prevent this. In their view, even our new proposal gave the Air Force too prominent a role. They held that it was enough to coordinate plans developed by the services. But we argued that the United States needed a single plan, integrated from the beginning by a joint planning staff.

The Secretary of Defense, Thomas Gates, became aware of this issue and made it known that he favored the approach being proposed by the Air Force. In the interest of advancing the cause of integrated planning, Secretary Gates asked his military assistant, Brig Gen George Brown (an Air Force officer who later became the Chairman of the JCS), to keep him informed of the status of this matter.

In an effort to get the ball rolling in our direction, we proposed a trial run. An ad hoc joint planning staff would convene in Omaha and develop a trial SIOP to test the concept. The Navy opposed this proposal, but the Secretary of Defense took affairs into his own hands and directed the JCS to conduct this trial. After a month or two, the ad hoc planning staff produced a first cut at a SIOP, but there was still considerable dissension between the Air Force and the other two services about a variety of issues.

There was a lot to argue about. The Air Force wanted the plan to take into account the possibility that poor weather would hamper or even prevent operations by carrier-based aircraft. The Navy and the Army pointed to problems regarding several issues having to do with Air Force forces. They argued that far fewer bombers than estimated by SAC would be likely to penetrate Soviet air space and reach their targets. They also considered that the Air Force was far too optimistic about the proportion of bombers that could be launched before Soviet ICBMs destroyed them on their bases. In addition, they argued that the Air Force was overestimating the hardness of Soviet industrial targets. They feared that this assumption would have the effect of raising the requirement for more nuclear forces and more nuclear weapons.

In an effort to force a resolution of these matters, the Air Force submitted a memorandum to the JCS, proposing that the Joint Chiefs and the Secretary of Defense convene in Omaha to review the draft plan that the ad hoc planning staff positioned there had developed. Predictably, the Navy and Army opposed this suggestion. Again, the Secretary of Defense intervened. He directed that the meeting be convened on December 18, 1960. In addition, he directed that the objective of the meeting be changed from “review the plan” to “approve the plan.” The Chief of Naval Operations was so strongly opposed to this wording that he asked to see President Dwight Eisenhower. The outcome of the meeting was predictable: The President sided with his Secretary of Defense.

The ad hoc staff had briefed the plan to the JCS in Washington several times in preparation for the meeting in Omaha. The entire issue was nearing some resolution, and it was certainly a top priority for all concerned. The Air Force intended the following chain of events: The draft plan would be briefed to all concerned (including Secretary Gates) on December 18 in Omaha. The Air Force would then recommend approval of the plan, with expected opposition from the Navy and the Army. The Air Force confidently expected that the secretary would decide in favor of approval, in light of the split, when the decision was put to him. Thus, the plan would be approved.

But General Gerhart wanted nothing left to chance. “How do we know the Secretary of Defense will rule in our favor?” he asked. “We must have a compelling argument to this end. You develop this argument,” he told me.

Stuck in traffic as I was going home that night, I pondered how we could make an airtight case for our position. Then it struck me that the right way to think about this was to view the SIOP as an OPLAN (which it was) and as a means for making the best use of the forces available—with not a hint as to any requirement to meet a set of objectives. The plan should be presented as a set of allocations, i.e., allocating this or that weapon to each designated ground zero (DGZ). What effects these strikes would have and whether these were adequate to meet the nation’s needs were important but separable questions. One could argue endlessly about such issues as degradation caused by adverse weather and darkness, the probability of our bombers penetrating Soviet airspace, the hardness of the targets, what portion of the Soviets’ industrial worth these strikes were likely to destroy, whether these percentages represent enough destruction or too much, and so forth. All these arguments tended to be inconclusive because no one had the experience of all-out nuclear war.

From the perspective of a SIOP—narrowly defined—all these arguments were beside the point. The task given to General Power and his planners was straightforward: Make best use of the forces we have. That meant that there was only one output: the allocation of weapons to DGZs. The Air Force should declare that the sole purpose of an OPLAN was to make optimal use of available forces, which our plan did. There would inevitably be debate about the likely effects of nuclear strikes and whether these strikes were adequate or not, but these were separate issues.

Once this position was staked out, the Air Force needed to show that the protests by the Navy and the Army concerned the likely effects and adequacy of nuclear strikes; they were not about the plan itself. To this end, I proposed that General Power hold a meeting of the ad hoc staff in Omaha. This meeting would include a two-star Navy admiral and a two-star Army general. The participants in that meeting would review all the items that were under protest by any of the services. This would surely be a long and rancorous session. At the end of this session, General Power was to inquire in a quiet manner whether the Navy admiral or the Army general had any recommendations with regard to the assignment of weapons to DGZs. If so, would they please submit those recommendations to him?

We hoped that the other services would make some recommendations, but not too many. General Power would implement all their recommended changes. At an appropriate time during the meeting on December 18, General Power would state that he had asked for recommendations on DGZs, that he had received certain recommendations from the Navy and certain recommendations from the Army, and that he had changed the plan to reflect all these recommendations. Therefore, he would declare that the plan itself was not in contest. The likely effects of these nuclear strikes were a topic for continued debate.

General Gerhart liked this gambit. He told the Air Force Chief of Staff, General White, about it and General White called General Power. Without revealing any details over the phone, General White said that he would send me to Omaha to explain the approach. I arrived at Offutt Air Force Base (AFB) late that same evening and was promptly whisked to General Power’s office. I explained to him the gambit that I had proposed to General Gerhart. The central point was that the Air Force be able to tell the Secretary of Defense on December 18 that there was no protest about the plan itself.

“It will work,” said General Power. “How many others know of this gambit?”

“Only General Gerhart and the Chief,” I replied.

“Then keep it that way.”

We kept everything under wraps until the meeting on December 18. On that day, all the ranking civilians and military officers were in attendance at SAC Headquarters in Omaha. In fact, General White, a four-star, just made the cut to sit in the first row. The agenda items for the meeting on December 18 all concerned various factors that could change the predicted outcome of the nuclear strikes. As we predicted, the participants engaged in a heated debate about these various factors. Just before lunch, General Power requested the floor. He stood in front of the group and spoke directly to the Secretary of Defense. He explained the difference between factors that affect the outcome of strikes and, on the other hand, the allocation of weapons to DGZs. He said that, with respect to various factors, there was much debate, but that with respect to the allocation of weapons to the DGZs, there was none. He stated that he had accepted all the changes proposed by the Navy admiral and the Army general about the allocation of weapons to DGZs.

Secretary Gates was more than impressed. He said, “General Power, if what you say is true, then this casts quite a different light on this matter.” He then extracted a grudging admission from the Navy admiral that he had indeed submitted five changes concerning DGZs and that all had been approved. He extracted a similar admission from the Army general. The secretary then closed the discussion. He said that, since there were no disagreements regarding the allocation of weapons to DGZs, the plan should be approved without further debate, without change, and today.

But the Chief of Naval Operations, ADM Arleigh Burke, was not quite finished. “Mr. Secretary,” he said, “I think that it would be rather awkward if the Congress came to know that you coerced the Joint Chiefs of Staff into approving the plan before it was officially submitted to the Joint Chiefs for approval.”

Oh my. I had overlooked this one important detail. The plan had been briefed to the JCS three times, and the JCS had a copy of the plan. But the plan had not been officially submitted to the JCS for approval. Any delay would give the Navy time to recover. All the principals were scheduled to depart the premises that afternoon and return to their respective bases. Just as we seemed to have closure, it was slipping away. There was no telling what would happen if the plan were not approved during the session in Omaha. The Navy would recover and now make some stern protest about the allocation to DGZs. I anticipated that General Gerhart would not be pleased with this outcome or with me.

Then Secretary Gates came to the rescue. “Admiral Burke,” he said, “you have a point to which I must react. The plan will be submitted to the Joint Chiefs officially today. You will have all night to consider it. I now amend my earlier statement. The Joint Chiefs will report to me in the morning at nine o’clock as to which members approve the plan and which members do not. If one member approves, I expect the matter to be presented to me for adjudication. I will surely find in favor of the member who has voted for approval.”

We were back on track.

The Navy admirals and Army generals protested about staying over in Omaha because they had appointments to keep. The secretary brushed the objections aside. “I expect to see each of you in this room tomorrow at nine,” he announced. Then he asked, “General Power, may we partake of your hospitality for one more night?”

General Power beamed. “Of course. There will be a reception in the Officer’s Club beginning at 1830 hours.”

At the reception that night, General Power took me by the arm and ushered me into the presence of General White. “General White, this is the man who made this happen,” he said. From that time forward, I was a protege of General Power, much to the benefit of my career in the Air Force.

The rest is history. In light of the secretary’s dramatic statements, the Navy saw no value in continued opposition. At the meeting on December 19, all three members of the JCS voted to approve the plan. Once the decision was made that a JSTPS would be established in Omaha for the purpose of developing a SIOP, the Navy and Army worked hard to make the SIOP a milestone in developing OPLANs. [*]

[* – Joint Secretariat, “Review of the Initial NSTL and SIOP,” note to the Joint Chiefs of Staff on JCS 2056/194, December 9, 1960. For more detail on the process of creating the first SIOP, see Headquarters, Strategic Air Command, “History and Research Division, History of the Joint Strategic Planning Staff: Background and Preparation of SIOP-62,” n.d.]

This SIOP affair was my crowning achievement as a colonel. It is a lesson that big things can be accomplished by diligent and persistent staff work—especially when the Secretary of Defense is on your side. The following underlined my doctrine: Be sure your position in the JCS is so compelling that, if there is a split, the Secretary of Defense will surely rule in your favor. In this case, the key to success was to conceive of a way to frame the debate such that arguments against our position were simply untenable. It was clear from the start that efforts by the Air Force to have the Navy’s Polaris fleet “chopped” to the commander of SAC were doomed. But the desired effect—greater coherence in the OPLAN for executing the forces of both services—could be achieved in a different, and more politically palatable, way. Who could argue legitimately against a joint planning staff, especially when it was made clear that the product of that staff—an OPLAN—was intended to make best use of the forces available?

In later years, SAC became simply Strategic Command (STRAT-COM), a joint combatant command analogous to U.S. Central Command or U.S. Pacific Command. And more often than not since then, the commander has been a Navy admiral with submarines, bombers, and intercontinental missiles under his control. This arrangement would have been unthinkable to all concerned in the early 1960s. Time changes many things, sometimes for the better.

In this episode, the Air Force gained its point. The result was beneficial not only for the Air Force but also for the country. Had the other services prevailed, the United States would have gone on planning Armageddon in a disjointed way. At best, planning by the individual services would have caused inefficiencies, invited redundancies, and made the nuclear deterrent less credible. At worst, such planning might have caused uncertainty and ragged decision-making in a time of crisis. Whatever parochial concerns may have motivated the Air Force to advocate a single integrated plan, it was clearly in the national interest. Finally, one might argue that the SIOP set a standard for jointness that eventually expanded to conventional operations, especially through the Goldwater-Nichols reform.

Pages 30 – 37
Defending the Planners of the SIOP

The time was the mid- to late 1960s. Gen John McConnell was Chief of Staff of the Air Force, and Dr. Harold Brown was Secretary of the Air Force. I was head of Air Force Studies and Analysis (AFSA). General McConnell called me to his office. He said that there was a big problem. Three analysts from the Office of the Secretary of Defense (OSD) had made a visit to SAC. They had been briefed on the SIOP and on the “planning factors” used in the preparation of this document. They had then written a report that was very critical of several of the planning factors used by the JSTPS in developing the SIOP.

These analysts had delivered their report to the Secretary of Defense, Robert McNamara. The report included a recommendation that a group be formed to thoroughly review how the JSTPS had developed the various planning factors and how these planning factors were applied to defining the SIOP.

General McConnell went on: “This is not just an effort to improve the SIOP,” he said. “These people have a hidden agenda. They have in mind that the ‘review group’ they are proposing will declare that the JSTPS at Omaha is inept and that this planning should be done by a group reporting directly to the Secretary of Defense.” The general added that he had been informed of the existence of this hidden agenda by a very reliable source. He went on to point out that events were moving rapidly. He and Dr. Brown had just returned from a meeting with the Secretary of Defense. Secretary McNamara had informed them that, in view of the report on his desk, he had little choice but to form the group to review the planning. But General McConnell saw this, and rightfully so, as a slippery slope. If the group for the review were formed, there was a strong likelihood that the responsibility for developing the SIOP would be taken away from the JSTPS. He was determined that this whole affair be stopped in its tracks.

General McConnell went on about the discussion between Dr. Brown, the Secretary of Defense, and himself. “While the secretary was explaining that he had no choice,” he said, “it suddenly occurred to me that I had a trump card to play. I recommended that the Secretary of Defense not establish the review group until General Kent has had a chance to review the critique by the people from OSD.”

The Secretary of Defense accepted this recommendation. He knew me well from my days in OSD under Dr. Brown. The general went on: “Your job is to show that the analysts at Omaha are just as sharp as the analysts from OSD. Show that there are serious flaws in the OSD report.”

At that point, Dr. Brown entered the room. “I presume General McConnell has told you of what happened,” he said. “I wish to emphasize—there is to be no whitewash. Whatever your findings may be, they must be able to stand up to critical review.”

Putting the two statements together, my marching orders were clear: I must show that SAC is right, that the OSD analysts are wrong, and the case must be airtight.

A quick look at the OSD report made it easy to believe General McConnell’s statement about the hidden agenda. The critique was wide ranging. It was obvious that the report’s authors were trying to establish a basis for taking the responsibility for developing the plan away from the JSTPS in Omaha. They had much more in mind than simply trying to make some improvements to the SIOP itself.

As noted in “The Advent of the SIOP” ( pp. 22-30), the SIOP, in effect, defines the allocation of various nuclear weapons among DGZs: Weapon number 22 goes to DGZ number 1; weapon number 23 goes to DGZ number 2; and so on. In rare instances, more than one weapon is assigned to the same DGZ. If you make a change in the SIOP, it is a matter of changing weapon number 45 from DGZ number 21 to DGZ number 28. No big deal. The marginal return of making this change is undoubtedly small and impossible to measure. The effect desired was deterrence: That is, we sought to convince the Soviets that it would be against their interests to launch an attack. [*]

[* – Some people thought the effect desired was to create unacceptable damage in a retaliatory attack. That may be so, but it was always clear to me that we were better served by keeping our focus on deterring a Soviet attack in the first place.]

In this context, a change in the allocation of particular weapons would result in little change in the degree to which we deterred—especially if the Soviets were unaware of this change, which would be considered Secret or Top Secret.

Now, back to the OSD critique. The analysts identified, as I recall, ten issues that they regarded as evidence of the JSTPS’s incompetence. One of these (item 4) caught my eye: “The weight of effort allocated to attack the Tallinn complex is ridiculous.” They actually used these words verbatim. This intemperate language stood out and made this item a likely candidate for rebuttal. If I could show that the planners in Omaha were about right in this allocation, we would have a leg up in tarnishing the report by the three OSD analysts.

The Defense Intelligence Agency (DIA) had for some time—up to five years—been reporting on worrisome activities by the Soviets near Tallinn, the capitol of Estonia. Their estimate was that the Soviets were installing an antiballistic missile (ABM) complex to shoot down U.S. missile reentry vehicles (RVs) as they made their way to targets in the Soviet Union. Activities had been observed at some 40 separate sites around Tallinn. DIA was not certain how many interceptors had been deployed (or were to be deployed) at each site and was not all certain of the effectiveness of each interceptor, but it did give a range: between 20- and 80-percent effective—whatever that meant.

The planners in Omaha, in the presence of these tentative assessments by DIA, assumed (1) that 15 interceptors had been (or might be) deployed at each of the 40 suspected sites and (2) that each interceptor had around a 65-percent probability of kill (P_k) given a launch. In the presence of these assumptions, the planners had allocated five RVs per site, for a total of 200 weapons, to suppress the Soviet ABM system. It was the number 200 that bothered the OSD analysts. It seemed like overkill, especially in view of the uncertainty surrounding the complex. Accordingly, they had labeled the allocation “ridiculous.”

In truth, at first blush it does seem like overkill (allocating five weapons per site)—if the lethality of each weapon is such that one is all that is required to destroy all the interceptors at one site. But DIA had stated that the probability of intercept might be as high as 80 percent (or words to that effect). The Soviets would use the interceptors at the site in self-defense, and there would be only a 20-percent probability (at worst) of each U.S. RV penetrating, as long as the ABM system is operating. So, it makes some sense to put a sizable number of weapons onto the Tallinn ABM system to ensure that the complex is destroyed and that the U.S. RVs attacking other DGZs are not intercepted. The question remains: What is the optimum number of RVs to commit to attack this complex?

I began to consider ways to quantify the value of suppressing this defense. Starting from the simple case of a single RV against a single ABM site, we can calculate (on an expected-value basis) that the RV would cause one interceptor to be launched in self-defense and destroy 2.8 of the remaining Soviet interceptors:

0.2 * (15 - 1) = 2.8 (The 0.2 comes from 1 – 0.8.)

Thus, 11.2 Soviet interceptors would remain.

If two RVs were allocated per site, 8.32 Soviet interceptors would remain:

0.8² * (15 - 2) = 8.32

So there is merit in allocating more than one RV per site. I saw that we should expand these calculations to determine the “optimum number of RVs per site.” My measure of optimum in this case was obvious: It is the number of RVs used in defense suppression that maximizes the number of RVs that penetrate the defense and proceed to attack other (non-ABM-related) targets on the territory of the Union of Soviet Socialist Republics (USSR).

A discussion with the planners in Omaha revealed that the weight of effort (200 RVs) expended against the Tallinn complex had been discussed—albeit very briefly—with the OSD analysts during their trip to Omaha. The SAC planners were somewhat amazed that the analysts had chosen the word ridiculous to characterize this allocation (200 RVs total). The SAC planners pointed out that, if they assumed a probability of intercept of around 65 percent, it took a little more than five weapons per site to attain a damage expectancy (DE) of 0.90 per site. That is, 0.65⁵ = 0.12, and 1 - 0.12 = 0.88 DE. And that was about the extent of the discussion on that item.

During their visit to SAC, the analysts from OSD did not challenge the number of sites (40) or the possible number of interceptors per site (15), mostly because the subject was not discussed in detail. Rather, they challenged the requirement for 0.90 DE. They opined that while the 0.90 DE might have been sacred to the planners in Omaha, it had no firm basis in policy or mathematical analysis.

I then undertook a simple analysis. The measure of merit was the number of RVs to penetrate the defense and reach DGZs in the USSR. Suppose that 1,000 RVs were involved in an attack. The planner would be willing to divert 200 of these RVs to defense suppression if, by doing so, the number of RVs available to attack other DGZs would increase.

The notional characteristics of the Tallinn ABM complex were as follows:

1. 40 surface-to-air missile (SAM) sites

2. 15 interceptors per site

3. for a total of 600 interceptors.

Table 1.1 reveals that, for the conditions stated, the optimum number of RVs per site in defense suppression is between four and five. This allocation maximizes the number of U.S. RVs that penetrate the Soviet defense, as shown in the far right column of the table.

So it turns out, fortuitously, that the planning factor of 0.90 DE and a probability of intercept of 0.65, as used by the SIOP planners, gave approximately the right answer, though perhaps not for entirely compelling reasons. But I was not obliged to defend the reason. The OSD analysts in their report to the Secretary of Defense had not directly challenged the use of a DE of 0.90. Rather, they had challenged only the number of RVs diverted. So with regard to item 4, OSD was wrong, and the planners in Omaha were right. I explained all this to General McConnell and, in turn, to Dr. Brown. Dr. Brown went through my calculations number by number and was satisfied that the logic was sound.

Dr. Brown then convened a meeting with the general, with the three OSD analysts who had written the report and myself in attendance. I started with a dissertation on the strategy to be followed when planning in the face of uncertainty. I noted that planners must make choices about the allocation of forces and that their choices must be predicated on judgments about a wide range of factors whose precise values are unknown. To be robust, a plan must be executable in the face of adverse circumstances. Hence, the correct strategy is to maximize the outcome for the case in which the factors in question are adverse. We should, in the vernacular, “maximize the min.” We should plan against the worst case. In this case, that meant planning for the case in which the probability of intercept (PI) for the Soviet ABM system was 80 percent. It would be imprudent to adopt a planning factor that maximized the outcome when the PI was 20 percent.

Notice that I had reduced the problem to whether the PI to use was 20 percent or 80 percent. This spread was derived from the DIA estimates. I announced then that allocating five RVs per site, a total of 200 RVs against 40 sites, was about right for the factor of 0.80. I added that simple math could be used to demonstrate this.

I had expected that rather than challenge my math, the analysts from OSD would challenge the assumptions about the structure of the complex I had assumed—especially that the number of interceptors per site was 15. If one assumed that fewer than 15 interceptors were present, the optimum number of RVs per site would be less than five. To my surprise, they chose not to challenge either my math or my assumptions. Instead, they declared that they were “not going to get in a numbers game” with me.

I replied, “You raised the issue of numbers. You stated that an allocation of 200 RVs (five per site and 40 sites) was ‘ridiculous.’ My math demonstrates that five is about right. Where is your analysis that shows that five is so far wrong that it warrants the characterization of ‘ridiculous’?”

At that point, Dr. Brown observed that my confrontational tone was not conducive to a productive discussion, whereupon General McConnell, to my surprise and chagrin, stated that he agreed with the secretary and summarily dismissed me from the room. As usual, the general was one step ahead. In about 30 minutes, he called me to his office. “I got you out of the room,” he said, “before they could change tactics and try to challenge your analysis. After you left, I made the point that, if they continued to pursue this matter, I would make the point that they, when faced with your analysis, refused to be drawn into a numbers game. I would certainly make this known to the Secretary of Defense—and anyone else who would listen.” The general went on: “While they did not agree to cease and desist, I think we have heard the last of this matter. They, of all people, should avoid making sweeping statements before they do their homework.”

And so it was. Their report withered on the vine. They no longer pushed the matter. Their choice of the word ridiculous was ill advised, and I had turned it into a fatal error.

In summary,

• In cases of this nature, it is almost always preferable to attack than to defend. It would have been very difficult for me to show convincingly that everything the JSTPS was doing was correct. At best, this approach would have resulted in a series of charges and countercharges with no conclusive resolution. But if I could discredit the report of the OSD team, we might convince Dr. Brown and Secretary McNamara to drop the affair.

• If you undertake to challenge or discredit a report, focus and dwell on the one item about which its authors are obviously wrong and be prepared to prove your point.

• Never use such polemical words as ridiculous in your own work unless the number you are challenging is very wrong and you can prove it. Even under those circumstances, find a less inflammatory word.

In the wake of this episode, I was elevated to a higher stature in General McConnell’s eyes. He would at times introduce me as his “junkyard dog.”

Pages 37 – 42 – Calculating the “SIOP Degrade”

In 1991, Maj Gen Robert Linhard was the Director of Plans and Resources at SAC in Omaha. I had retired from the Air Force and was working at the RAND Corporation. There had been some discussion in Washington about deploying an active defense to counter any attack against the United States by a third country (that is, one other than the Soviet Union) with nuclear-armed ballistic missiles. Such a deployment would not be consistent with the ABM treaty of 1972. Rather than withdraw from the treaty, the United States was considering a concept that would allow the Soviets to deploy a similar defense along their southern borders to guard against an attack by their neighbors to the south.

General Linhard wanted to know how such a defense, if allowed, would affect the United States’ ability to execute the SIOP. In the terms he used, he sought to understand the “SIOP degrade.” To this end, he invited a group of analysts to convene in Omaha to address this question. The invitees included analysts from the Los Alamos, Livermore, and Sandia National Laboratories; two analysts from the RAND Corporation (Dean Wilkening and myself); one representative from the U.S. Arms Control and Disarmament Agency (ACDA); someone from the Department of Defense (DoD) Office of Research and Engineering (DDR&E); and others.

The meeting opened promptly at 0830. General Linhard stated his question and problem. The three analysts from the national labs set forth their approach to answering the question. After further questions and discussions, we finally disbanded for lunch. After lunch, General Linhard opened with a statement: “What I learned this morning is that, if I gave each of you a million dollars, after six months of ‘computer crunching,’ you could provide some tentative answers.” They all nodded affirmatively.

At two o’clock it was my turn. I opened by saying, “General Linhard, I am prepared to provide you considerable insight as to the implications of such a Soviet defense, this afternoon and for free. First, the bottom line: The degrade to the SIOP will be minimal (something like 10 percent), provided (1) that you employ decoys based on the latest decoy technology and (2) that you forbid the Soviets to deploy any forward engagement radar in the northern part of their country (a radar to control the engagement of a Soviet interceptor engaging U.S. RVs).”

To scope the problem,

1. Assume that each side is allowed no more than 200 interceptors.

2. Assume that each interceptor has a P_k (for a given engagement) of 0.7, for a total kill potential of 140.

3. Assume that the attack by Blue is 1,000 RVs against 1,000 targets, one RV per target.

4. Assume that 1,000 targets contain people and industrial facilities amounting to 1,000 units of “worth”—the proper choice of a scaling factor can make this true. Note that, for convenience, one unit of worth is an erdel (erdel is a fabricated word and means nothing).

5. Assume that the distribution of worth (erdels) among the 1,000 targets obeys the distribution according to the equation by the renowned economist Vilfredo Pareto, in which

V_cum = [ (n / N)^0.5 ] * w

where V_cum is the cumulative value, n is the stated number of targets, N is the total number of targets in the set, and w is the worth in the total set. The 1/2 is the exponent Pareto derived, which of course indicates the square root. It follows that one-half of the total value is in the first one-fourth of the targets.

Note that I was not breaking new ground with regard to the distribution according to Pareto. Many analysts have used this distribution, including the assignment of one-half as the value of the exponent.

6. Assume that the Soviets are not allowed to deploy an “engagement” radar to the north of the tier of provinces in the south of their empire.

7. Without an engagement radar to the north, they cannot engage in “threat-tube sorting.” That is, they cannot determine the intended target DGZ of each attacking RV and thus employ their 200 interceptors against the 200 RVs that are destined for the 200 most lucrative targets. Note that the defense has great leverage if it has a control system that provides this capability. Without it, the defender has to engage the incoming RVs without regard to the value of the target that each RV is intended to destroy.

8. Also, by not allowing the Soviets to deploy an engagement radar to the north, we assume that they cannot discriminate between decoys and RVs in the attack by Blue.

For the base case (no defense), the defense of course saves no worth (zero erdels), and there is no degrade to the SIOP.

Now take the case of a defense with a kill potential of 140 (200 * 0.7). There are 1,000 RVs versus 1,000 targets, and each target contains (on the average) one erdel of worth. Thus, the defense “saves” 140 *1.0, or 140 erdels. This amounts to a 14-percent degrade.

For the next case, add 400 decoys to the attack. The decoys are, perhaps, Mylar balloons and very light. Assume that we must allocate 20 weapon spaces to employ the 400 decoys—a rate of 20 to one. The 980 remaining weapons attack 980 targets. These 980 targets contain 990 erdels:

1,000 * SQRT(980 / 1000) = 990

Now, with 400 decoys (and 1,380 total objects), the expected erdels saved per object killed by the defense has been reduced from 1.0 to 0.717:

990 / 1,380 = 0.717

assuming that the Soviet defense system is unable to distinguish between RVs and decoys.

Then, the erdels saved by killing 140 objects is equal to 100. To this number, we must add the erdels saved by replacing 20 weapons with decoys. This number is 10. So, the total erdels saved by the defense is 110, and the degrade is 11 percent.

If the technology of lightweight decoys permits 40 decoys per weapon space allocated rather than 20, the degrade will be a little less. The defense now “saves” less per object killed, 0.556 versus 0.717, as was the case for 400 decoys. Now, the defense saves 78 erdels (140 * 0.56)—plus the 10 erdels in the 20 targets not attacked—for a total of 88 erdels, a degrade of slightly under 9 percent for the case of 800 decoys.

On the other hand, if the Soviets can do threat-tube sorting, the problem takes on a different complexion. With threat-tube sorting (and no credible decoys), the Soviets would allocate their 200 interceptors to attack (selectively) the 200 RVs that are directed toward the 200 most lucrative targets. These 200 RVs put at risk nearly 45 percent of the total value of the system of 1,000 targets. The expected value saved by the defense per RV destroyed is now 2.25, and a kill potential of 140 saves 315 erdels. Now, the degrade is a whopping 32 percent.

We can see the powerful reasons for outlawing forward engagement radars: (1) There is no threat-tube sorting, and (2) there is no discrimination between RVs and decoys.

At this point, some of the other analysts could see their prospects for substantial future work slipping away and proclaimed that the problem could not reliably be reduced to such a simple calculus. Both General Linhard and the gentleman from ACDA thought otherwise, and they stated that they now had adequate insight to inform some key policy decisions:

1. Outlaw forward engagement radar (or radars).

2. Hold the number of interceptors the Soviets were allowed to a low number—a few hundred.

3. Pursue a vigorous program to develop lightweight decoys.

In the presence of these three provisos, the problem of SIOP degrade was minimal and presented no compelling argument against negotiating with the Soviets to allow both countries to deploy a limited defense to counter the threat of attack by third countries.

The key here is to scope the problem and address the key assumptions in a manner distinct from a focus on “computer crunching.” In other words, just sit back and think. Often, doing this can provide a basis for calculations that, while quite straightforward, yield new insights into the most important aspects of the problem.

For reasons external to those examined here, the whole concept of each side deploying a limited defense was lost in the turmoil surrounding the collapse of the Soviet Union and the United States’ decision, in 2001, to withdraw from the ABM treaty of 1972.

Pages 43 – 50
Limiting Damage to the United States

Early in 1964, a study group was formed under the auspices of OSD’s Office of Systems Analysis, headed by Dr. Alain Enthoven, to “look at” a wide range of issues relating to U.S. nuclear forces and strategy. Dr. Frank Trinkl was to run this group for Dr. Enthoven. At the time, I was a brigadier general working in DDR&E for Dr. Harold Brown.

As someone with experience in nuclear forces, I was asked to serve on this group. I declined. In my view, there was next to no chance that this group would succeed in providing insights useful for informing decisions on the critical issues of the day. Several things were wrong:

1. There were far too many issues to address (18 in all).

2. There were far too many people involved. Some had been called to the Pentagon on temporary duty and had no place to work.

3. There was no hint as to the analytical approach that would be used to gain insight on any particular issue.

That I had declined to serve was made known to Dr. Brown, my boss. He called me into his office. I told him of my misgivings about the way the effort was being run. He rather agreed but directed that I serve anyway and do my best to make it a useful endeavor. To me, this amounted to a mandate to try to reshape the effort.

My first recommendation had to do with “grouping.” There were several issues on the study group’s agenda that had something to do with limiting the damage to the United States that might be caused by nuclear attacks. Why not bundle them together under one large effort called “limiting damage” and conduct a comprehensive analysis of the prospects of significantly limiting the damage to the United States from a Soviet attack and, additionally, define the proper allocation of resources to the various means for doing so? These means would include the following:

1. counterforce operations

2. active defense against ballistic missiles

3. active defense against bombers

4. passive measures (civil defense).

Dr. Trinkl opined that such a study would require a major effort that he did not have the time to undertake. I pointed out that I had not introduced any new or additional items. I had merely laid out an analytical approach to gain insight into a major and encompassing issue. Indeed, one of the items he listed was just that: “limiting damage.” Absent a comprehensive study, we were doing nothing more than “looking at” several items without providing much-needed insight on any one of them.

Dr. Trinkl repeated that he was reluctant, given his time constraints, to undertake the substantial effort I had proposed. He had ample reason for this reluctance. Dr. Trinkl held fast, and so I departed the group. I made sure that Dr. Brown heard of my decision (and of my reasons for it) from me and not from someone else. He listened and nodded. “You are right. Do the study yourself,” he said. Ouch. I was now embarking on an effort that would consume most of my time and energy for the next year or so.

The first thing I did was recruit some good people. My choice from the Air Force was Maj Jasper Welch. I knew him from earlier days. I asked a Navy admiral I knew for the name of a Navy officer to serve on my team for this huge effort. He mentioned a Captain Paolucci. “But you don’t want him. He is something of a maverick. He is very smart but can be difficult to work with.”

I thought for a moment. “Send him down,” I said. “I would like to talk to him.”

Capt Dominic Paolucci opened the interview by interviewing me. “What am I getting into?” he asked. “I don’t want to waste my time working for another flag officer who doesn’t know squat about analysis.” I explained my approach and showed him some notional plots. Also, I added, we would try to provide insight as to the allocation among the various means (players) based on marginal return.

Captain Paolucci was all ears and he began to get interested. “I would like to serve,” he said. Recruiting Captain Paolucci was one of the best moves I made. He was a brilliant analyst and became one of my best friends.

After several weeks, Major Welch, Captain Paolucci, and Captain Niemela of the Army, using mostly notional data, produced a first cut, “proof of principle” assessment of the primary options for limiting damage from a Soviet nuclear attack. We presented this preliminary briefing to Dr. Brown as an update about work in progress. To our surprise, he then showed these charts to Secretary McNamara, who in turn directed that there be a comprehensive effort with participation by each service and, as well, the staff of the Civil Defense Office.

Suddenly, the stature of the study went up dramatically. This was both good news and bad news: The representative from each service was to be a two-star general; I was only a one-star. There was also an effort to establish a “steering group,” ostensibly to advise my group on the conduct of the study. I deftly avoided such a group. We were now engaged in a comprehensive study that was to have many twists and turns.

One of the first things to do was to get a better cut at the database for population and manufacturing value added (MVA). A contractor laboriously went through U.S. census data and defined circles in rank order as to their “worth,” according to population and their value in terms of MVA. The location of the first two circles was not surprising: They were in Manhattan. Only one circle in California made the first ten. At that time, California did not account for as large a portion of the U.S. population or gross domestic product as it does today. Moreover, even though the state ranked high in population, it was quite dispersed. The circles, as I recall, had a radius of five miles— approximately the lethal radius of a 1-megaton bomb.

So that was our construct. The United States would strive to defend the worth in each of these circles, and the Soviets would attempt to destroy it. We assumed that the Soviets would attack where the worth destroyed per missile expended would be greatest. These are the most lucrative DGZs. Needless to say, this database became close-hold. The circles were ordered in the first instance by population.

We also had to consider the fatalities among the population from the radioactive fallout from Soviet counterforce attacks on our nuclear forces, the Minuteman silos in the Dakotas being the prime example. If the wind came from the northwest, the debris from these attacks would cause considerable fallout over Chicago and surrounding areas. To determine the number of fatalities from this fallout, we had to make assumptions about the proportion of the population that would seek refuge in fallout shelters, the proportion that would survive if they did, and the proportion that would survive without shelters.

We addressed the Soviet bombers’ attack as follows: We would deploy enough active air defense in sufficient numbers of interceptors so that the Soviets would see no advantage in acquiring and deploying bombers as part of their attacking force.

Obviously, there were many assumptions to discuss (and argue) with the participants from the services, each a two-star with a “dog in the fight.” Finally, we had some answers. We were careful to state that the study was not intended to predict the outcome of an attack with any precision. Rather, it was intended to provide insight into two issues:

1. What are the prospects for limiting damage to the United States from a determined and adaptive Soviet attack? [*]

[* – We were careful to construct the analysis so that the Soviet attack could be adjusted to whatever defensive measures the United States put in place. Failing to do this would have led to results that overstated the value of the defenses.]

2. What is the proper (best) allocation of resources among the various “players” involved in limiting damage:

a. civil defense with passive measures, such as fallout shelters in many areas and blast shelters in the largest cities

b. active defenses against arriving Soviet ballistic missiles with Nike-X

c. counterforce attacks with Minuteman missiles against Soviet ICBM bases, submarine-launched ballistic missile (SLBM) ports, and bomber bases

d. antisubmarine warfare operations with U.S. submarines against Soviet SLBMs at sea

e. active defense with U.S. interceptors versus Soviet bombers?

We arrived at these allocations based on a marginal return. [*]

[* – Unfortunately, I have misplaced many of these graphs over the years, and they are too complicated to reproduce now.]

We finally published the study. The essence of the study is detailed in Figure 2.1.

The figure shows, for each utility level,

1. the total resources the United States must spend to achieve that utility level in the presence of the stated Soviet attack

2. the appropriate allocation of resources to each player

a. civil defense

b. ballistic missile defense (BMD)

c. strategic offensive forces

d. antisubmarine warfare

e. bomber defense

3. The ratio between U.S. expenditures to limit damage and Soviet expenditures to create damage as required to maintain the stated utility level.

For example, the figure shows that, to achieve the level of 70-percent survival of the U.S. population (against a stated Soviet deployment of ICBMs and SLBMs), the United States would have to spend a total of $28 billion. If the Soviets reacted to our deployments to limit damage by deploying more ICBMs and SLBMs, the United States would be obliged to spend more money to stay at the 70-percent level. The exchange ratio (the amount the United States would have to spend to limit damage compared to the amount the Soviets would have to spend to create damage) was adverse to the United States. At the margin, it was always cheaper to create damage than to limit damage. The graph shows exchange ratios of 0.8 and 1.7, respectively, at the 70-percent and 90-percent survival levels. We realized (belatedly) that the published numbers, which reflected the official exchange rate between the ruble and the dollar, understated the exchange ratio. When the values were revised on the basis that the costs to the Soviets to purchase ICBMs and SLBMs were comparable to our own costs, the ratio was more like 2:1 at the 70-percent survival level. At the 90-percent level, the ratio was more adverse—probably 6:1. The charts actually briefed to Secretary McNamara reflected the revised figures.

The secretary observed that this was a race that we probably would not win and should avoid. He noted that it would be difficult indeed to stay the course with a strategy that aimed to limit damage. The detractors would proclaim that, with 70 percent surviving, there would be upwards of 60 million dead.

The secretary went on: Instead of seeking unilaterally to limit damage, we should undertake to negotiate a treaty with the Soviet Union to curtail the deployment of nationwide defenses. This could set the stage for agreements to control the deployment of offensive forces. Needless to say, this was a statement of great and lasting strategic importance.

So my efforts to gain insight into options for limiting damage had an unexpected ending. Rather than reordering priorities for investments among approaches to limiting damage, the study resulted in a rather fundamental change in policy that led the administration to more or less abandon efforts to limit damage in a meaningful way.

“Limiting damage” did not appear as a stated strategic objective in the draft presidential memorandum (DPM) issued in 1964.

Pages 50 – 54
Limiting Damage: Allocation of Resources

As stated earlier, the study on limiting damage addressed the question of the proper allocation of resources among the various “players”:

1. Civil defense

2. BMD

3. strategic offensive forces to engage in counterforce operations against Soviet ICBMs in silos and Soviet SLBMs in port

4. antisubmarine warfare forces to conduct counterforce operations against Soviet submarines at sea

5. air defense forces (both area and terminal defenses) to intercept Soviet bombers.

The allocations, in general, were based on marginal return. All the players operate with diminishing returns, and we have the optimum allocation when all players are operating with the same marginal return, that is, when a stated increment of funds yields the same increase in the measure of merit (U.S. population surviving) regardless of the player to which this increment is granted.

We determined the optimum allocations using graphs. We plotted graphs with the measure of merit on the ordinate and money allocated to a particular player on the abscissa. Once the graph was plotted, we could take a ruler and determine the slope of the line at any particular level of expenditures allocated to that particular player. The slope, by definition, is the ratio of the change in the measure of merit to the change in investment, where the measure of merit was either the population surviving or the MVA surviving (see Figure 2.2).

We created such a plot for each of the five players. It was, to say the least, a complex, iterative, and laborious approach. But the team persevered, and finally, we arrived at a point at which, with reasonable confidence, we could state the optimum allocation of resources among the various players.

For example, we concluded that, for the stated size of attack, the United States had to spend a total of $28 billion for 70 percent of the population to survive. That $28 billion should be allocated as follows:

• $5.2 billion to civil defense

• $12.8 billion to active BMD

• approximately $6 billion to strategic offensive forces

• approximately $1 billion to antisubmarine warfare

• approximately $3 billion for bomber defense.

Today, with high-speed computers, the optimum allocation to each player (at various levels of total expenditures) could be determined more quickly but not necessarily more reasonably. We conducted many excursions to test the proposition that allocations other than the ones we stated as optimum would yield worse results in terms of the measure of merit (U.S. population surviving).

One challenge of note regarding allocations stemmed from a group of people on contract at the Civil Defense Agency. They were charged with evaluating the concept of constructing blast shelters in the largest cities. We chose not to implement this concept even though, at the higher level of expenditures on BMD, the marginal return was equal to or better than the slope for BMD for population surviving. What disqualified blast shelters from our final comparisons was that, unlike every other player, they offered no protection to the economic infrastructure of the United States. In short, when the measure of merit is MVA surviving, blast shelters for the people were not considered a player. Since BMD operates on both of the measures of merit, the nod goes to BMD. Not surprisingly, the Civil Defense Agency contractors were not convinced and continued to challenge our decision not to recommend that this concept be implemented. As it turned out, none of the concepts were implemented. In this sense, one could say that the operation (the analysis) was a success, but the patient (limiting damage) died.

Another challenge regarding allocations came from the representative that the Air Force assigned to my study group. We had stated that, in each of our optimum allocations, the United States should provide the capability to employ one effective RV against each Soviet ICBM silo. This was an increase from the planning factor of zero set forth in the DPM, developed by Dr. Enthoven. This was not a new issue. Dr. Enthoven argued, with some traction, that the United States, as a matter of policy, would not engage in a first strike. Thus, in any conflict, most of the Soviet ICBMs would already have been launched before our RVs arrived. This meant that our counterforce attacks against these empty silos would have minimal effect in reducing the number of Soviet ICBMs arriving at U.S. targets.

Our analysis on limiting damage revealed that counterforce operations with Minuteman RVs (one Minuteman per Soviet silo) were a viable option, even if only 20 percent of those silos still housed an ICBM when the Minuteman missiles arrived. We stated that we could not predict with any certainty how a future exchange might unfold.

Still, in the presence of this uncertainty, we should adopt a planning factor of one RV per silo.

Dr. Brown and Dr. Enthoven were impressed by this argument, and it was agreed that DoD would revise the planning factor up from zero to one. But the Air Force representative, a two-star general, wanted more. He argued for a planning factor of two RVs per Soviet silo. To bolster his position, he argued that the P_k of a Minuteman RV against a Soviet silo was 0.9. We had used 0.6 for the Pk, and with good reason.

I pointed out to the Air Force general that the marginal return of the second Minuteman RV would be minimal, as the Pk of the first RV was 0.9. This was so because 0.9 times 0.1 (the residual value of the silo that had already received an RV) is 0.09. If you choose to argue for two, I said, then you should allege that the P_k is 0.5, since x times 1 - x is at a maximum when x = 0.5. Still, the general held to the position that, if we would just use 0.9, the stage would be set for a planning factor of two, when in fact alleging that the P_k was 0.9 destroyed any argument for the second weapon. As a matter of fact, a P_k of 0.6 was not far off from 0.5 in terms of the marginal return of the second weapon: It was 0.25 for a Pk of 0.5 and 0.24 for a P_k of 0.6. Nevertheless, the Air Force general refused to grasp the disconnect in his train of logic.

He had other complaints as well. He bundled them together and proclaimed to the Chief of Staff of the Air Force (then Gen Curtis LeMay) that I was “selling the Air Force down the river.”

General LeMay convened a meeting, the stated purpose of which was to determine whether or not I was selling the Air Force down the river. General LeMay began the meeting by stating his intention to demote me back to colonel. At that meeting, the two-star presented his case—including the issue that he had submitted that the P_k was 0.9 and I had ignored his input and used 0.6. In my response, I dwelt on this point. Fortunately, General LeMay saw the disconnect in the logic presented by the two-star, namely that a P_k of 0.9 did not set the stage for a planning factor of two. In fact, it was the other way around. I pointed out that I had convinced Dr. Brown and Dr. Enthoven of the merit of allocating one Minuteman RV to each Soviet ICBM silo—up from zero. I pointed out that the issue of using 0.6 for the P_k as a planning factor was hardly any evidence that I was selling the Air Force down the river. As a matter of fact, it was quite the contrary. At this point, General LeMay abruptly ended the discussion and departed the room. The crisis abated, and I heard no further word on this matter.

Pages 54-56
Helping with DPMs

During the tenure of Secretary Robert McNamara, there was, as I mentioned above, a document known as the “draft presidential memorandum.” Actually, it was none of these things. It was not a draft; it did not go to the President; and it was not a memorandum. Nevertheless, it was a defining document whose focus was on defense strategy and the associated means. It defined objectives and the strategy we would use to achieve them. It also discussed the means and capabilities we intended to use to implement our strategy and the forces we intended to field to gain these operational capabilities—all in considerable detail. It also contained programming data, such as the amount of money allocated to each program element in the budget to field these forces.

All the above was administered by the Planning, Programming, and Budgeting System (PPBS), then run by Dr. Charles Hitch, formerly of RAND. Dr. Enthoven, also formerly of RAND and who reported to Dr. Hitch, was responsible for writing the portion of the DPM that dealt with strategic nuclear forces and strategy. The overall construct was a marvel of clear thinking. The section by Dr. Enthoven was invariably logical, insightful, correct, and short—on the order of 30 pages. I was heavily involved in the preparation of the strategic nuclear section simply because of Dr. Brown. Dr. Hitch had to gain approval from Dr. Brown, and Dr. Brown consulted me on what was to be stated in this section.

I address the matter of the DPM not to chronicle my part (which was secondary) but rather to compare this document and to contrast its development with the approach used today. Documents in the mid-1990s, such as the Report on the Bottom-Up Review and the Secretary of Defense’s annual reports, compare favorably. But not so today; the Strategic Planning Guidance recently issued is 300 pages long, and it fails, in my judgment, to define in clear terms either our strategy or the means we intend to use to implement that strategy.

Dr. Hitch, Dr. Brown, and the secretary himself would review the DPM, item by item and line by line. In addition, the Chairman of the JCS and the service chiefs conducted their own review. Their comments were taken seriously, but they held no power of veto.

One example in particular illustrates the clarity of the document. The document stated the administration’s paramount strategic objectives: to deter strategic nuclear attack by the Soviets on the United States; to limit damage from such an attack if one should take place; to deter the Soviets from launching an all-out attack on Western Europe; and, if necessary, to halt such an invasion as far forward as possible (i.e., near the intra-German border).

Our strategy with respect to the first objective, deterring the Soviets, was to have the operational capability to inflict severe damage on the USSR. We gained this capability by deploying SLBMs on submarines at sea, by deploying weapons on bombers on quick alert at bases in the United States, and by deploying nuclear weapons on ICBMs in hardened silos in the United States.

Prior to the completion of the damage-limiting study, our strategy with respect to limiting damage was to increase our operational capability to conduct counterforce operations against Soviet ICBMs in silos and SLBMs on submarines in ports, to deploy an active defense against incoming Soviet RVs and bombers, and to construct a stated number of fallout shelters (but no blast shelters). From this point, having laid out the basic defense strategy, the DPM dealt mostly with programming data.

The DPM was indeed a defining document in shaping the direction the United States was to take—both with respect to strategy and with respect to means. The DPM reflected the decisions of the time and was revised annually until it was eliminated by Secretary Melvin Laird during the Nixon administration. The DPM of 1963 included some definitive guidance as to measures the United States would take to limit damage. In the DPM of 1964, these statements were quietly absent, a change stemming from the study on damage-limiting I had conducted. [*] In light of that study, Secretary McNamara had decided against allocating large amounts of resources to the various means of achieving that objective.

[* – For more detail, see “Limiting Damage to the United States,” pp. 53—50.]

Over the years, I had, from time to time, several arguments with Dr. Enthoven, some of which were quite heated. On the other hand, I always applauded his skill and insight in drafting the DPMs. We would do well to emulate his approach and construct today.

Pages 137 – 139
The Short-Range Attack Missile Affair

Another program that incurred unneeded turmoil as a result of the misuse of the term requirement was the program to develop a shortrange attack missile (SRAM) for the B-52 strategic bomber. The program began in the mid-1960s. The concept was to equip the B-52 with a nuclear-armed missile to attack a Soviet SA-3 air-defense site with impunity. The SRAM would be fired at the site from a distance beyond that at which the SA-3’s radar at the site could detect and track the B-52.

The calculation for the “required” range was straightforward. At 23 nmi, an aircraft flying at 400 feet above ground level cannot be detected and tracked by a radar on a 90-foot tower with a grazing angle of 6 degrees to the horizon. At a lesser range, the radar could detect and track the aircraft. So the 23 nmi range was indeed an operational requirement. If the range of the missile was less than 23 nmi, the SA-3 battery could not be attacked with impunity, and the concept would be null and void.

A development program was commenced. By about the late 1960s, the SRAM was approaching initial operational capability (IOC). I was in AFSA at the time. A colonel from another office brought in a report for my coordination. It was an annual report to Congress as to the status of the acquisition programs the Air Force was conducting, system by system. For each system, there was an item called “Requirement.” To my surprise, the “requirement” listed in this document for the SRAM was that the missile have a maximum range of 37 nmi. I had been involved in framing the concept for the SRAM some years before and remembered that calculations determined the maximum range to be 23 nmi. After some investigation, we learned that the switch to 37 nmi came from a calculation by an engineer at Boeing, the prime contractor for the program, who determined that the system they were building would have “thrust” and “drag” such that it should be expected to fly 37 nmi. I pointed out that this longer-range figure was a statement about expected performance, and, harkening back to my earlier experiences, I added that there was some danger in putting it down as an “operational requirement.” The colonel agreed in good faith to change the draft he was coordinating to state that the requirement was 23 nmi, but his superior, a general running the directorate for “requirements,” overruled him. “Why put down 23 nmi when we are on contract for 37?” he reasoned. They did not inform me of this decision.

A year went by. A new annual report went to Congress, one that I did not see before it was dispatched. In this report, the Air Force listed next to the entry for the SRAM that the range “requirement” was now 33 nmi. In the year since the previous report, the engineers at Boeing had encountered some unexpected problems, and the thrust they were able to achieve went down, while the drag went up. The system now under development did not now meet the previously stated requirement of 37 nmi. A staffer for Rep. Samuel Stratton (D-NY) spotted the change from 37 to 33 and informed the congressman: Congressman Stratton went public, stating that the missile cannot meet the required range and should be cancelled. The matter was referred to the U.S. General Accounting Office (GAO) for investigation.

By now, the whole affair gained the attention of the Chief of Staff of the Air Force. He gave me the job of stopping the GAO report. I was not successful; the GAO stated that 37 nmi (the required range) could not be attained. In fact, according to the GAO, even the 33 nmi range was in doubt. The office recommended the program be cancelled. Being accountants, they knew the difference between two numbers but not the relevance of either. In due time, Congressman Stratton held hearings before a rump session of the House Armed Services Committee. At those hearings I explained that the original calculation of the required range (23 nmi) was based on such operational matters as the need to be able to attack Soviet SAMs with impunity. I stated that, if the expected range became as low as 25 nmi, the Air Force on its own would cancel the program.

Congressman Stratton replied that my presentation helped to clarify the distinction between a genuine “operational requirement” and “expected performance.” In light of this distinction, he expressed bewilderment at the fact that the Air Force had ever stated that the requirement was 37 nmi and asked me how such a thing could happen. In response, I blurted out, “Because we have our share of people who sometimes do not think straight.” To this, the congressman replied, “If that be the case, then I expect you to take remedial action.” “Done,” I said.

Following the hearing, I reported to the chief that Congressman Stratton no longer considered the SRAM an issue. However, I noted that the congressman enjoined us to take action so that colonels who do not always think straight have no part in preparing documents that are sent to Congress. The chief dutifully gave a brief statement at the next staff meeting about the danger of confusing operational requirements with expected performance, and the colonel was transferred out of the Pentagon. Even so, the practice of blurring this distinction continues and is still alive and dangerous today.

Pages 144 – 146
The Minuteman III

From the very beginning of the program, the number three was bandied about as the number of RVs for the front end of the new Minuteman III. But Dr. Brown felt that we needed better insight into this matter. To this end, he tasked the Minuteman SPO to conduct an analysis to provide better insight into the matter of defining the front end of the Minuteman III. In due time, they came back with an analysis that was more complicated and extensive than it was insightful. Dr. Brown sent them back to the drawing board. Weeks later, they again appeared in his office. Again he rejected their analysis—not so much because it gave the wrong answer, but because it was neither insightful nor compelling.

Upon their departure, he directed me to go out to Norton AFB, California, the home of the Minuteman SPO, and stay there until they produced an analysis that he could use to defend the number of RVs per missile on the Minuteman III—whatever the number was. I pointed out that there were two separable issues: (1) creating an analysis that would meet Dr. Brown’s standards and (2) having the SPO produce that analysis. I blithely announced I could provide the analysis, but I was doubtful of my ability to get the SPO to do an analysis that passed his muster.

“All right,” he replied. “What is it?”

“Sir, it is a matter of elimination. The candidates range from zero to five. By inspection we can eliminate zero.”

He failed to see the humor in this.

“It can’t be one. We already have that covered. It can’t be four or five, as calculations by our engineers show that, if the missile is to boost a PBV with four RVs 5,500 nmi, the weight of each RV is so constrained and the nuclear warhead is so small that the yield goes down the drain. That leaves us with two or three.”

To choose between two or three, we must decide on the value of the “scaling factor.” The scaling factor of one-half controls the trade-off between number of RVs and the yield of each when attacking an area containing “soft” targets. The scaling factor normally used was two-thirds, but I had shown previously that it was more like one-half. The derivation of the scaling factor for nuclear weapons is summarized in “The Trade-Offs Between Numbers, Yield, and CEP in Hard-Target Kill,” pp. 226-229.

When using the scaling factor of one-half, two RVs or three RVs were about a tie as far as destructive power was concerned: That is, a Minuteman missile with two RVs had about the same destructive power as a missile armed with three lower-yield RVs. But the three RVs were obviously preferred in terms of countering an ABM system. First, more RVs meant that there would be more objects for the Soviet ABM system to counter. And, second, if in the future it was determined that decoys would need to be deployed, the decoys appropriate for simulating the smaller warhead would presumably be lighter than those called for by a two-RV design, so there could be more of them for the space or weight available. Thus, three was better. Dr. Brown concurred.

Later, Dr. Brown revisited the matter of the Minuteman III. He wanted insight as to the number of missiles to deploy and the number of RVs to deploy on each missile. Since he, in effect, had asked two questions, this indicated that the number of RVs per missile was not entirely settled.

After several false starts, I arrived at the following construct: The requirement is to have some number of RVs (x) survive a Soviet attack. The Soviet attack is defined by how many Soviet RVs are assigned to the attack of the Minuteman force and the probability of an RV killing a Minuteman silo once assigned.

Empirically, from engineering data, I took that whatever a one-RV missile may cost, a four-RV missile costs twice as much, and a nine-RV missile costs three times as much, a scaling by the square root. If you deploy very large missiles, the cost per RV deployed goes down, but the number of aimpoints also goes down, and the expected fraction of RVs surviving the attack also goes down. So there must be an optimum number of RVs per silo (missile)—optimum in the sense of having a stated number of RVs surviving a stated Soviet attack and doing so at least cost.

After several futile attempts, I solved the problem. There was an “optimum” survival percentage, and that number was, interestingly enough, [1 / e] (that is, one over the Napierian e, or 0.37). So, the number of RVs to deploy was the “requirement” times e, or times 2.718. On the other hand, the number of silos to deploy was such that the quantity (1 - Pk)^n = 0.37, where n is the number of Soviet RVs assigned to the attack on the Minuteman divided by the number of Minuteman silos deployed.

Thus, if the requirement is that 600 RVs are to survive, then we should deploy

600 * 2.718 = 1,631 RVs

If the threat is 800 Soviet RVs with a Pk of 0.5 each, then deploy 552 silos, since and

[800 / 552] = 1.45

and

0.51^1.45 = 0.37 = [1 / e]

I showed this math to Dr. Brown, and he announced the analysis was clever—maybe too clever. We would get different answers if we changed the requirement or the threat. That is so. But reasonable cuts at these two parameters provided a rationale for what we were planning, namely 550 silos and 1,650 RVs.

I recite this example to underline the idea that, if the analyst works hard enough, and long enough, and clearly enough, he or she will eventually arrive at a simple analytical solution to a complex problem.

Page 158 – 160
Gaining Insight as to the Vulnerability of Submarines on Patrol

In the early 1970s, I served on a group directed by the Secretary of Defense to develop recommendations about the basing of our strategic nuclear forces: ICBMs in hardened silos, SLBMs in stealthy submarines, and bombers on alert at SAC bases. Among other matters, we were to examine and render findings as to the vulnerability of each mode of basing.

The Navy promoted the idea that missiles based on submarines at sea were quite invulnerable. To support this, they presented an analysis that had to do with Soviet attack submarines on patrol attempting to detect, track, and kill a Polaris submarine on station in the Norwegian Sea. According to their analysis—which used an expected-values approach to calculate the probability that a Soviet attack submarine would encounter, detect, establish track, maintain track, and kill the SSBN—the Soviet attack submarines were “never” successful.

Following the meeting in which this analysis was presented, I discussed the matter with Jasper Welch, by now a colonel. He suggested that we use their same inputs for values in the kill chain but determine the outcomes stochastically. That is, rather than multiply all the expected probabilities together, we should use a Monte Carlo simulation to determine the outcome (yes or no) of each event in the kill chain:

1. Soviet submarine detects Polaris: yes or no

2. Polaris submarine detects Soviet submarine and takes evasive action: yes or no

3. If “no” above, Soviet submarine tracks Polaris submarine: yes or no

4. Polaris submarine finally detects Soviet submarine and takes evasive action: yes or no

5. If “no” above, Soviet submarine engages and kills Polaris submarine: yes or no.

We ran many engagements. Once in a while, there was an unbroken series of “yes” answers for the Soviet submarine, and thus a “kill”— a successful engagement. We used the Navy’s own probabilities in our Monte Carlo simulation to determine “yes” or “no” for each event in the sequence. The point here is that, if you try to achieve an event many times, you will finally succeed once in a while. Treating the outcomes as expected values tends to obscure this point.

In less than two weeks, amazingly, Colonel Welch had developed a computer model that ran these engagements in a stochastic manner.

Once in a while, according to this analysis, the Soviet submarine was successful in detecting and engaging the Polaris submarine. Colonel Welch presented his work to the group. I introduced the item not so much to make the point that Polaris subs are vulnerable but rather to make the point that we should be cautious about moving everything to sea, as the Navy seemed to want.

Not surprisingly, the Navy representative on the working group was not pleased with our effort, our analysis, or our results. In fact, the next day, I was summoned by the Secretary of the Air Force, who had been called by the Secretary of the Navy, to explain why I was engaging in analysis of Navy systems. I felt that the Navy’s objections to our work were way out of order. It was well known that the Navy was not shy about analyzing the survivability of Air Force systems, and that, in fact, they had maintained a group of analysts who episodically issued analyses underlining the vulnerability of ICBMs in hardened silos and bombers on alert.

The Navy’s objections notwithstanding, we had made our point:

The members of the group from OSD were impressed by the analysis and had a better understanding about the importance of avoiding a policy of putting all the nation’s eggs in one basket.

Part of the Navy saw some merit to our approach. Colonel Welch related to me a month or so later that the Navy had decided that it should do its analysis stochastically and had quietly asked Colonel Welch to explain his computer program.

Even so, the Navy did not dismantle its analytical “hit team” and continued to render analyses of the vulnerabilities of missiles in hardened silos and bombers on alert. [*] But, for the moment, our analyses had stemmed the tide.

[* – My annoyance at the Navy’s practice of issuing such analyses was another reason I went after their SLBM survivability numbers. I had hoped to make them desist from criticizing our systems ]

Page 199 – 201
Assessing the Effectiveness of Bomber Attacks

In the late 1950s, when I was head of the Weapons Division on the Air Staff, we focused intently on the question of to what extent our bombers (mainly B-47s) could penetrate Soviet air defenses in a retaliatory attack (after a Soviet attack on the United States). Later, the mainstay of this retaliatory attack was ballistic missiles—SLBMs or ICBMs— but the burden was on bombers in those days.

Gen Glen Martin, the deputy director of plans, revisited the issue surrounding bombers over and over. He wanted the answer to many “what-ifs”: What if a certain fraction of our bombers was not launched in time for “safe escape”? What if only a certain number penetrated the outer zone (corridor) of Soviet air defenses? (He asked the same question for each zone; there were four in all.) What is the merit of attacking more targets (DGZs) that are “shallow” versus attacking highervalue targets that are “deep,” by which we incur more losses of bombers before they reach their targets (release points)? And so on, and so on.

General Martin had directed these questions to a group of analysts in another division. He was not satisfied with the results of their efforts and informed me that, from now on, I would be in charge of answering his questions and would have oversight of this group for this purpose. General Martin inundated us with questions. He was asking questions much faster than we could generate answers. The “turn time” for answering each question was typically days—if not many days. The team worked long hours and on weekends. But the harder it worked, the more behind it was. Answers to questions spawned more questions.

Something had to give:

Either we had to reduce General Martin’s appetite for information, or we had to come up with an easier and faster way to generate answers. Knowing General Martin, the latter approach seemed more tractable. We decided to turn to what we called in those days a large-scale computer. The approach was as follows:

1. There will be a card (a punch card) for each individual nuclear weapon in the SIOP.

2. This card names the weapon (i.e., number so and so).

3. This card also names the delivery platform that carries it (i.e., what particular bomber, ICBM, or SLBM).

4. It indicates where the delivery platform is based.

5. It specifies the DGZ the weapon is to attack.

6. It designates the corridors (zones) of defense that the carrier has to penetrate to reach the stated DGZ.

Then, we would use Monte Carlo simulations to determine

1. which weapons (by tail number) were safely launched from their home bases

2. which weapons (by tail number) actually penetrated the defense corridors they were required to penetrate to reach their DGZs

3. the actual ground zero (AGZ) of each weapon.

We compared these AGZs to listings in a database that catalogued the locations of people, MVA, and industrial facilities and military facilities in the Soviet Union. We then could announce that, given a weapon of this yield detonating at this particular location, so many people would be killed and so many industrial (or military) facilities would be destroyed.

We needed help to do the programming. Computers were, in those days, not user-friendly. There were only a few people who were versed in the art of programming. We let a contract to IBM to provide four programmers. IBM was eager to help and welcomed the opportunity to show how its magic machines could be used to inform important decisions.

In about four weeks, we were up and running. I had convinced General Martin to cease and desist with his questions until we could develop this new tool. Fortunately, we did not have to start from scratch as far as the database was concerned. Now, the “turn time” was three hours, maybe less. Since the model was stochastic, we were obliged to make several trials for each “what-if.” We settled on ten trials and then printed the results for the “median” trial. The printout was in two different formats: one for the generals and a more-complicated one for analysts. To be sure that we could stay ahead of the game of questions and answers, I gradually revealed that the turn time was now a matter of hours.

General Martin was very impressed with this effort. In fact, he made known our approach to the planners at SAC, and I lost the services of two of the experts from IBM.

To repeat, given a different input as to the probability of penetrating a stated defense zone, the computer sorted the “cards.” In this “sorting,” some weapons were taken out of the game. For those that penetrated to the release point, we used Monte Carlo to determine the AGZs. Incidentally, our high-speed computers consisted of tapes and wheels in a console taller than me, and it took awhile for each trial. Now, a computer, once programmed, could run each trial in the blink of an eye.

Page 202 – 212
Assessing the Effectiveness of ABM Deployments

My first assignment after my year at the Center for International Affairs at Harvard was as Military Assistant to the Director of Research and Engineering. This assignment was about as “good as you get.” My boss, Dr. Harold Brown, was a man of considerable intellect. The Secretary of Defense, Robert McNamara, relied heavily on the advice of Dr. Brown on an array of issues—many beyond the direct purview of DDR&E.

As it turned out, Dr. Brown, in turn, listened to others whom he had learned to trust. About a month after I arrived, the Army brought in a milestone study about the effectiveness of the Nike-X in terms of limiting damage to the United States from a Soviet attack with nucleararmed ICBMs. The measure of merit used in this study was the proportion of the U.S. population surviving a stated Soviet attack, with and without a stated deployment of Nike-X interceptors.

The Nike-X interceptor was nuclear armed. It was “unguided” after launch and was launched to a point in space according to a tracking radar at the site. The P_k given an engagement was calculated as 0.50, hopefully greater. The range of the interceptor was stated as 8 nmi, so an interceptor at a site could defend an area with a radius of 8 miles.

The Army analysis was according to the following construct:

1. They had obtained “from the JCS” a copy of the RSIOP—a Soviet version of their attack on the United States, like the U.S. SIOP.

2. The Soviets would attack each DGZ with two missiles.

3. Each Soviet missile would employ one warhead and nine decoys.

4. Our radar (the Nike-X radar) could not distinguish between decoys and RVs.

5. The Nike-X battery defending the DGZ would fire two interceptors at each object; so, 20 Nike-X missiles would be fired for every Soviet ICBM that was engaged. Each Nike site had 40 missiles ready to fire, so each site could engage two Soviet ICBMs.

6. Given the above, the Army then calculated the fatalities from the Soviet attack—with and without the Nike-X deployment of 40 interceptors per site.

The reduction in fatalities afforded by the Nike-X deployment was shown to be significant. I was getting bad vibes about the analysis from the start. I remembered the advice of Dr. Schelling, who taught “The Strategy of Conflict” in a course at Harvard. He cautioned us to think carefully about which side has the “last move.” The construct the Army used implied that the United States had the last move: The Soviets defined their attack; that is, they chose their DGZs and assigned two missiles to each one. The United States was made aware of the character of the attack and responded accordingly, by deploying 40 interceptors at each DGZ.

The reality, as I saw it, was quite different: The Soviets would observe the deployment of the Nike-X battery by battery and tailor their attack accordingly. They were not obliged to attack the defended areas defined by our deployment. Rather, they could attack undefended areas if they chose to do so. Surely, then, the Soviets had the last move. And if so, the difference in U.S. fatalities with and without the Nike-X would surely be much less than shown by the Army analysis.

I was the new man on the block, so I held my tongue while the Army was briefing Dr. Brown. However, when the briefing was over, I followed Dr. Eugene Fubini back to this office. Dr. Fubini was Dr. Brown’s principal deputy and a man with a razor-sharp mind. I had hardly finished explaining the “last move” problem when Dr. Fubini stood up abruptly: “Of course, of course,” he said. “Come with me to Dr. Brown’s office.” There, the argument was repeated. “You are absolutely correct,” Dr. Brown said. “Call the Army back here.”

Dr. Brown explained his concern to the Army. “We cannot use this analysis as a basis for informing our decision about deploying Nike-X,” he stated. This demanded a new question: How do you go about deciding where and how many interceptors will be deployed when the Soviets have the last move? The Army and its contractor, California’s Stanford Research Institute, were back to the drawing board.

A day later, the other shoe fell. Dr. Brown directed me to “think through this problem.” Ugh.

After wrestling with the problem for a week or so, I had made some progress, but not much. I now realized there was a related issue: the firing doctrine at the battery. How many interceptors do you employ per “object” in the attack if you do not know how many objects are yet to appear? You must balance “leakage” against “exhaustion.” From the standpoint of leakage, you employ many interceptors per object. But the more you employ, the sooner your supply of interceptors is exhausted. The analysis by the Army had no problem in this regard. The Soviets always fired two missiles per DGZ, and knowing this, we always engaged all 20 of the objects. But if the Soviets attacked some of the DGZs with three missiles instead of two, the defenders would go from modest success to certain failure. If the attack is two RVs per DGZ, there is a 0.56 probability that no RV (warhead) will detonate [*]; if three, there is an absolute certainty that one will because the defender has no more missiles with which to engage the third RV.

[* -- This is so because 0.5^2 = 0.25 and 0.75^2 = 0.56.]

Then luck came my way. Someone told me that two analysts at Bell Telephone Laboratories had issued a paper on this very subject. They were Dr. Robert Prim and Dr. Thornton Read. I called Dr. Prim. He was delighted to come to the Pentagon and show me their work.

Their construct was absolutely elegant. The primary elements were as follows:

1.    The defense charges a price, p, for a stated target.

2.    The price charged, p, is directly proportional to the “worth,” W, of the target.

3.    A defense is defined by the ratio of W to p. They called this ratio lambda, λ.

4.    If the p of a target (a DGZ) is one, no interceptors are required.

5.    If p is two, then one interceptor against the first RV is required. If the Pk of the interceptor is 0.50, the Soviets gain one-half of the worth of the target with the first RV and 1.0 × one-half with the second RV.

Absolutely elegant.

I now turned to the task of calculating the “effectiveness” of a stated Nike-X deployment based on the Prim-Read theory. Given that the P_k given an engagement was a certain value and setting aside the issue of decoys for the moment, I devised an expression that gave the number of interceptors required as a function of (1) the price you charge and (2) the P_k given an engagement. In it,

I = [ ln( p!) / -ln(1 – Pk) ]

where I is the number of interceptors. For p = 4 and P_k = 0.5, this yields a total of 4.59 interceptors to charge a price of four RVs. If the price is four RVs,

1. There is a one in four probability of penetration for the first RV; two interceptors are required.

2. There is a one in three probability of penetration for the second RV; 1.59 interceptors are required.

3. There is a one in two probability of penetration for the third RV; one interceptor is required.

4. There is a 1.0 probability of penetration for the fourth RV; 0 interceptors are required.

The total is 4.59 interceptors, just as the formula said. You will note that the “expected return” for each RV is the same:

RV1: (1/4) * 1.0 = 0.25

RV2 = (1/3) * (3/4) = 0.25

RV3 = (1/2) * (2/4) = 0.25

RV4 = 1.0 * (1/4) = 0.25

To make it easier to do “what-ifs,” I used the formula to plot the number of interceptors required as a function of price charged for a range of P_k values (see Figure 6∙1 for an example).

With this “analog computer,” calculating the number of interceptors required became a simple matter of reading values off of a graph.

Once you understand the construct, the rest is easy:

1. Postulate λ, the ratio of W to p. The smaller the value chosen for λ, the larger the price you must charge for a target of a given worth.

2. Decide how many DGZ areas you intend to defend.

3. Calculate how many interceptors are required for each DGZ. This number depends on the λ you have chosen, the worth of that particular DGZ, the P_k of the interceptor, and the number of objects (RVs and decoys) in each missile. In the example above, since each Soviet missile carried ten objects (one RV and nine decoys) and the Nike-X radar could not distinguish between RVs and decoys, the total number of interceptors required to charge a price of four is 4.59 * 10 = 45.9.

Now make a plot of the worth destroyed as a function of the number of Soviet missiles in the attack. For this, I put the number of Soviet missiles on the abscissa and “worth destroyed” for the defended areas on the ordinate. This is a straight line starting at the origin and at a slope of λ. The line ends at the total worth contained in the “defended areas.” If the Soviet attack includes the “undefended areas,” tack on the worth destroyed per missile expended in the undefended areas.

If you have defended all “areas” whose worth is equal to or greater than λ, the marginal return of worth for each Soviet missile expended is exactly λ at the juncture of the lines of defended and undefended areas.

An example of such a plot is shown in Figure 6.2.

The figure reflects calculations of worth destroyed as a function of the size of a Soviet attack for two cases: (1) no defense and (2) a defense that allocates interceptors according to the logic of Prim-Read. Both cases in this somewhat notional example assume a target set containing 1,600 DGZ areas and a total population of 200 million. The distribution of the total population among the 1,600 DGZ areas is assumed to obey a Pareto distribution with an exponent of one-half. That is,

W_CUM = 200 * (n / 1600)^0.5

or

W_CUM = 5 * SQRT(n)

where W_CUM is the cumulative worth and n is the number of targets over which “worth” has been accumulated. In the no-defense case, this relationship is also what defines worth destroyed as Soviet missiles are traded one-for-one with DGZ areas, starting with the most lucrative areas to maximize the “expected return” of each missile.

In the case of the Prim-Read defense, λ has been set at 0.1. Thus, we want to defend DGZ areas up to the point where the “expected return” of Soviet missiles is 0.1 million per missile. The next issue that arises is how many targets we should defend. This question can be addressed target by target, but I find it convenient to work with sets of targets instead. Since cumulative worth obeys a Pareto distribution with an exponent of one-half, we can define 40 sets (40 being the square root of 1,600) with the following properties:

1. The number of targets in the kth set is equal to 2k - 1. The first set contains the most valuable target; the second set contains the next three most valuable targets; the third set contains the next five most valuable targets; and so on.

2. The number of targets through the kth set is equal to k². There are four targets in the first two sets, nine targets in the first three sets, 16 targets in the first four sets, and so on.

3. All sets are equal in value. Although each set contains a different number of targets, the value of the first set

5 * SQRT(1) = 5

is the same as the value of the 40th set

[5 * SQRT(40^2) ] - [5 * SQRT(39^2) ] = 5

and every set in between. This is simply the total worth of the target set divided by the number of sets, which is the square root of the total number of targets. In this example, each set contains a worth of 5 million.

Since λ = 0.1 and since each set has a worth of 5 million, the defense should charge a price of 50 for each set to be defended. If there are more than 50 targets in a set, it does not need to be defended, since the expected return per missile in that set would be less than 0.1. The 25th set contains 49 targets (2 × 25 - 1 = 49). The 26th set contains 51 targets (2 × 26 - 1 = 51). The expected return per missile expended in the 25th set is slightly greater than 0.1, and it is slightly less than 0.1 in the 26th set. Thus, we defend 25 sets containing 625 targets with a population of 125 million. The remaining 15 sets (975 targets with a population of 75 million) are left undefended.

In the absence of a defense, the Soviets could destroy half the total worth of the target set by attacking the 400 most lucrative areas with 400 missiles. In the presence of a Prim-Read defense with λ = 0.1, the Soviets would need 1,000 missiles to inflict the same level of damage. Moreover, only defended areas would be targeted, since attacking these DGZs brings the highest “expected returns” for Soviet RVs and the attack is not large enough to expect to destroy the total worth in the defended areas (a population of 125 million). Undefended areas would be left alone unless the Soviet attacks involved more than 1,250 missiles.

It is worth noting that, while the defense significantly raises the price of achieving a specified level of damage, it is not an inexpensive proposition for the defender. For example, the third set contains five targets with a population of 5 million. The population of an average target within the set is 1 million. Since λ = 0.1, the defender should charge a price of 10 for the average target, which would require 21.8 interceptors for the average target. Thus, roughly 109 interceptors are required to defend the set of five targets if there are no decoys or if the radar can distinguish between RVs and decoys. If we assume, as before, that each Soviet missile in the attack carries one warhead and nine decoys and that the Nike-X radar cannot discriminate between RVs and decoys, roughly 1,090 interceptors would be required to defend the five DGZ areas in the set. Around 11,000 interceptors would be required to cover the 625 defended DGZ areas.

Dr. Brown was quite impressed. He directed that the Army use this construct. They accepted the construct, but it took the Army’s contractor a while before he could get the computer to do his bidding. Accordingly, I spent a good deal of time doing “what-ifs” using a spreadsheet I had developed and a Friden calculator.

Doing “what-if” calculations on a target-by-target basis would have been extremely time consuming. After wrestling with the problem for a bit, I realized that I could reduce my workload substantially if I used a Pareto distribution with an exponent of one-half to represent the distribution of worth among U.S. targets. I could then take advantage of the properties of the distribution to define equal-value sets of targets. I placed the most valuable target in the first set, the next three most valuable targets in the second set, the next five most valuable targets in the third set, and so on. The total number of sets was equal to the square root of the total number of targets. For each set I calculated the number of interceptors required to charge the desired price for an average target and then multiplied by the number of targets in the set to determine the total number of interceptors required to defend the set. The set-based approach dramatically reduced the number of “rows” required on a spreadsheet. It was far easier and much faster to have to work only with, for example, 40 sets instead of 1,600 individual targets. My “turn time” was about two or three hours.

With a high-speed computer to run the spreadsheet, the turn time could have been reduced to minutes. Accordingly, Dr. Brown directed the Army to instruct its contractor to do the “what-ifs” by running my spreadsheet with a high-speed computer. Following several unsuccessful attempts by the Army’s contractor, Dr. Brown turned to the Institute for Defense Analyses (IDA) to do the “what-ifs.” IDA’s project leader then made a serious mistake. Instead of running the spreadsheet, he chose to render a critique of the whole calculus. He let a contract to this end to a professor of mathematics at New York University. This professor’s finding: While Colonel Kent’s methodology may be appealing on some counts, it has no firm foundation mathematically.

The professor offered a construct that had two features: (1) It would not work in terms of deriving the relationship among size of attack, interceptors deployed, and worth destroyed, and (2) it gave wrong answers for items you could calculate. He had made a careless error in signs: He had a plus when it should have been a minus.

Both Bob Prim and Thornton Read were infuriated at this report. After all, it was their construct that the professor had criticized. Dr. Prim went directly to Dr. Brown. IDA was again directed to do the “what-ifs,” but with no changes as to the spreadsheets.

The Prim-Read theory still stands as an exceptional piece of work. Analysts over the years have amended the calculus so as to account for noninteger solutions. (The defense cannot fire 1.59 interceptors at an object.) But the basic construct still stands. It was my luck to come across it and use if to good effect.

And so it was that my involvement in the calculus of the contribution of active defense led to thinking about the contribution of other means of limiting damage to the United States from attack by nuclear-armed ICBMs: (1) counterforce operations, (2) active defense, and (3) passive measures. This led to the defining study, “Limiting Damage to the United States,” pp. 43-50.

Page 226 -230
The Trade-Offs Between Numbers, Yield, and CEP in Hard-Target Kill

From time to time (in fact, too often according to my tastes), the AFSC program office for ballistic missiles would present to the Air Staff proposals to modify and modernize the Minuteman missile. During my tenure in AFSA, the Ballistic Missile Office (BMO) often sought to increase the HTK capability of the Minuteman force. HTK was defined as the number of Soviet missile silos that could be attacked by the Minuteman force with a damage expectancy of 0.9.

HTK can be increased in the following ways:

• increasing the yield of the warhead, which entailed a newer design and the use of more enriched uranium in each warhead

• decreasing the CEP using a new and improved guidance system

• simply increasing the number of missiles deployed

• pursuing some combination of all three approaches.

I was generally skeptical of these proposals. In the first place, the argument for seeking more HTK was not compelling: Even if the United States were to launch a first strike, the relationship between increased HTK and significantly reduced damage to the United States was obscure. The second problem with the proposals had to do with the “packaging” of increased yield, improved accuracy, and greater numbers of warheads into a single proposal. BMO’s presentations about the contribution of each component to the overall measure of merit were always opaque.

To gain insight into these matters, I developed the following construct:

• The measure of merit is the number of Soviet silos attacked at a stated DE of 0.9.

• The probability that a Soviet silo will survive an attack by one U.S. RV, Ps , is given by

  Ps = 0.5^(LR^2 / C^2)

where LR is the lethal radius of the warhead, and C is the CEP of the RV.

• Lethal radius is related to the yield of the warhead, y, by

LR = λy^(1/3)

where λ is hardness of the silo.

• The hardness of the Soviet silo, λ, was denoted as the lethal radius of a single 1-megaton warhead. Thus, if a 1-megaton warhead had a lethal radius of 1,000 feet, an 8-megaton warhead would have an LR of 2,000 feet, since

³√(8) = 2

Given these conditions, the Ps of a Soviet silo from an attack by a single warhead was

Ps = 0.5^[ [λ^2 * y^(2/3)] / C^2]

Then, the Ps from n RVs against one silo would be

Ps = 0.5^[ [n λ^2 * y^(2/3)] / C^2]

If we demand a DE of 0.9, it means that the exponent for 0.5 in the equations must have the value of 3.32 (since 0.5^3.32 = 0.10).

Now we can gain some insight as to the relative contributions to the total HTK of each of the measures proposed:

• If you increase the yield of a warhead from 250 kt to 400 kt (one of BMO’s proposals), you will increase the value of the exponent by a factor of 1.37, the calculation being:

(400 / 250)^(2/3) = 1.37

• If you decrease the CEP by a factor of 1.2, you will increase the value of the exponent by a factor of 1.44, that is,
  
  (1.2)^2 = 1.44.

• You can gain the same increase in the value of the exponent (namely, a factor of 1.44) by deploying 1.44 times as many deliverable RVs.

The first two measures (increasing the yield and decreasing the CEP) resulted in a combined increase of HTK by a factor of 1.97. In their presentation to the Air Staff, the BMO staff packaged the two measures together and stated that their proposal would “leverage the force by almost double.” But they were silent about the relative contribution of each measure. The total bill for their proposed program was, even by BMO’s own estimates, quite large, and I believed that their estimates were grossly understated.

General Ryan asked me to evaluate BMO’s proposal. In my briefing to him I confirmed that increasing the yield of the Minuteman warhead from 250 to 400 kt and reducing the CEP by a factor of 1.2 would, indeed, double the HTK of the Minuteman force. But I went on to point out that the majority of this increase in HTK stemmed from improving the guidance system, while the great majority of the cost of BMO’s proposal came from developing and deploying a new RV. This was so because many launches of the missile with the new warhead would be needed to test the RV’s ballistic properties and because large amounts of enriched uranium would have to be procured. According to my figures, the ratio of costs (of the new RV versus the new guidance system) was six to one. Based on my analysis, the Chief of Staff and the Secretary of the Air Force approved a program to develop the new guidance system for the Minuteman, but they did not approve a program for the new RV.

The above demonstrates that simple constructs can provide insight and reliably inform decisions about whether or not to proceed to implement some concept being proposed.

This episode was but one of a continuing battle between some members of the Air Staff and the ballistic missile program office. BMO continually sought funding to proceed with the development of a new RV for Minuteman, and I prevailed on the issue throughout my tenure in AFSA. However, subsequent to my departure from AFSA, a new Chief of Staff did grant approval to proceed with a new RV. The program was hardly a success—a large cost overrun occurred, and the resulting RV failed to meet its expected performance specification with respect to its yield.

The Trade-Off with “Soft” Area Targets

Now examine the trade-off between yield and numbers for the case of attacking “soft” area targets. Such targets include industrial infrastructure and unhardened military targets. In this case, the trade-off between numbers and yield is not so obvious. If the area occupied by the target is very large compared to the lethal area of one weapon, then the trade-off is the same:

(n*y)^(2/3)

where n is the number of RVs and y is the yield of each.

But that is seldom the case; industrial facilities are generally built not in large, circular clusters but rather more on a line (e.g., along a river or railroad within a valley). If the facilities lie in a line whose width is less than the diameter of the lethal area of the weapon, then we can, by math, announce the trade-off:

(n*y)^(1/3)

The one-third term in the exponent results from the fact that some of the weapon’s effects are expended outside of the target area. We have now bounded the problem: The exponent of y is somewhere between 0.33 and 0.67.

One might be tempted to take the arithmetic mean between these two values—0.5—but there would be a hue and cry about mathematical inelegance. So I devised a more complicated method for arriving at the answer. Specifically, I devised a chart with “Soviet value destroyed” on the ordinate and “number of weapons” on the abscissa and, for a family of curves, yields of 100, 200, 500, 1,000, and 2,000 kt. It took some effort to construct the lines, but finally we had the chart. Obviously, all lines (one line for each yield) started at zero and were concave downward—reflecting the fact that, as you went to more and more weapons on the abscissa, you were attacking targets of less and less value and thus for diminishing returns. Obviously, the 1-megaton lines rose more rapidly than the lines for lesser yields.

Now the trick: Draw a horizontal line from some place midway up the ordinate. This is a line of constant value destroyed. For example, we note that, for the line of 200 kt, it took 800 weapons to achieve this level of damage, and that, for the 1-megaton line, it took only 360 weapons. Now find the value of z so that 0.200^z × 800 is equal to 1^z × 360. The answer is z = 0.5. That is, when the exponent is 0.5, the square root of 0.200 × 800 = 358—close enough. Obviously, I took some other numbers and did not get the same value for z for all pairs. But the average value for the exponent was about 0.5—maybe a little less.

I now have stored in my mind that the exponent to use in determining the trade between numbers of weapons (n) and the yield of each (y) is Y^0.5 for the case of soft area targets. As we saw earlier, this relationship comes in handy in gaining insight into problems such as determining the number of RVs on the front end of the Minuteman III missile (see pp. 144—146).