The Last Warrior

by Andrew F. Krepinevich

  Schelling predicted the book would become a classic in the field, and it did. For senior US policy makers who had lived through the trauma of Pearl Harbor and who later confronted the prospect of a Soviet nuclear surprise attack, Wohlstetter’s book had a profound effect. Pearl Harbor was awarded the Bancroft Prize, and Wohlstetter would later be awarded the Medal of Freedom—the nation’s highest civilian honor—for her “great contribution to the security of the United States.”75

  Wohlstetter’s book was an early instance of how Marshall’s mentoring led to seminal work in the field of security studies. Given his innate modesty and inclination to avoid the limelight, this “hidden hand” approach suited Marshall. Over time other leading scholars and senior policy makers would benefit from his wise counsel, and acknowledge their intellectual debts to him.

  While Marshall was heavily occupied in the SOC, a new analyst arrived at RAND: Joseph Loftus. Although Marshall’s development as an analyst and strategist was certainly influenced by Goldhamer, Hitch, Kahn, Wohlstetter, Digby, and others at RAND during the 1950s, Loftus unquestionably had the greatest impact on the young strategist’s thinking about Soviet decision-making and the evolution of their nuclear forces. Perhaps even more important, Loftus indirectly contributed to Marshall’s growing conviction that a fundamental assumption of much contemporary analysis—that rivals could be counted on to behave “rationally”—was seriously at odds with their actual behavior.

  Prior to Pearl Harbor, Loftus had completed the course work and research for a doctoral dissertation on bank capitalization at Johns Hopkins University. But he never wrote his dissertation. A few months after the Japanese attack Loftus joined the US Navy, serving in Panama, the Aleutian Islands, and the Ryukyu Islands. After the war, he took a job as an assistant economics professor at American University, where he taught and conducted research until mid-1950. With grant money for his research running out, he then joined the Air Target Division of Air Force intelligence as a civilian, and spent the next four years tracking the USSR’s emerging nuclear program.

  Soon after Loftus joined RAND he and Marshall found themselves spending long hours together discussing “The Next Ten Years,” Soviet nuclear developments, and the Air Force’s strategic warning problem. Their talks quickly evolved into an effort to understand Soviet organizational behavior. Their collaboration lasted until the early 1960s when Loftus was forced to retire for medical reasons.

  In early 1955 Loftus, Marshall, and a RAND colleague, Robert Belzer, were asked to help the Air Force establish a worldwide network of centers to provide early warning of a Soviet nuclear surprise attack on the United States. With preventive war off the table the United States had little choice but to try to achieve strategic warning and avoid a “nuclear Pearl Harbor.” Some three years earlier, Marshall and Digby had gone to Wiesbaden to work on this problem, but the Air Force had been unwilling to grant them access to highly sensitive communications intelligence (COMINT) that could have provided critical insights regarding Soviet plans and decisions. This time Marshall, Loftus, and Belzer all received COMINT clearances, enabling Loftus to share with Marshall the detailed knowledge about the USSR’s nuclear program he had gained during his four years in Air Force intelligence. Once back in Santa Monica they were also able to keep abreast of ongoing Soviet developments by visiting the COMINT facility at March Air Force Base in nearby Riverside, California. At the time, only a handful of RAND staffers had access to this rich source of intelligence on Soviet nuclear developments. COMINT clearances gave Loftus and Marshall far greater insight into this area than nearly all of their RAND colleagues.*

  As their collaboration deepened, Loftus and Marshall began to realize how wide of the mark some of RAND’s forecasts about the USSR’s future strategic forces were. In one instance in 1954, a RAND research project attempted to predict where Soviet Long Range Aviation (LRA) bomber bases would be located. At that time US intelligence projections of Soviet nuclear forces were not yet being made on any consistent or regular basis. Forecasts might look out three years, or ten, or some number of years in between. In the absence of consistent projections of Soviet nuclear forces, RAND analysts had to make their own. Doing so required them to put themselves in the Soviet’s position. RAND researchers had to decide how they would position LRA’s bomber forces if they were responsible for them.

  Many RAND analysts assumed that the Soviets were supremely rational planners—that their major decisions about military forces were carefully calculated to pose the greatest threat to the United States in general, and to SAC in particular. This construct had great appeal at RAND and elsewhere in the US national security establishment because it allowed analysts to simplify their assumptions about Soviet behavior. By assuming the Soviets would behave rationally (at least as viewed from a US perspective), one could forecast future Soviet strategic forces without having to delve into the history, propensities, strengths and weaknesses, rigidities, military doctrines, operational methods, and organizational complexities of the Soviet state.

  Adopting this approach, the research on the LRA-basing issue concluded that the Soviets would locate their bomber bases in western Siberia along the Trans-Siberian Railway on a path from Omsk in the west to Krasnoyarsk in the east.76 This conclusion was intended to reflect how the Soviets would address the basing problem. It was based on the theory that “rational” Soviet decision makers would choose to make LRA’s bases as difficult as possible for SAC bombers to attack by locating them deep in the USSR’s interior.

  Loftus and Marshall knew from COMINT that this forecast was dead wrong. Soviet bomber bases were mostly dotted along the USSR’s periphery, their locations having been chosen during the early days of military aviation. At that time aircraft ranges were modest compared to what they later became, and basing LRA on the USSR northern periphery shortened the distances to US targets. The dilemma for Loftus and Marshall was that they could not share what they knew about LRA and Soviet nuclear programs with RAND colleagues who lacked COMINT clearances. The best they could do was to suggest alternative—or “nonrational”—ways of looking at the LRA-basing issue. Marshall and Loftus pointed out that most of the SAC bomber bases were on the periphery of the United States, where they had been since the early days of US military aviation. Might not this be the case with the Soviet Union’s? And if LRA’s mission was the atomic attack of the US homeland, then wouldn’t the limited ranges of early LRA bombers also argue for basing them on the USSR’s periphery? RAND analysts largely rejected these arguments as not fitting their rational-actor model of Soviet decision-making.

  There was, however, a larger strategic planning issue beyond this single RAND study. Early in their collaboration the more Loftus and Marshall looked into what the Soviets were actually doing as compared to what RAND and US intelligence analysts were forecasting, the more they began to realize that the forecasts relying on the rational-actor model often did not fit the facts. This realization led the pair to begin not only questioning the validity of the rational-actor model of Soviet decision-making, but to start looking for better models. The question they ultimately asked themselves was: Whose behavior and decisions are US analysts trying to forecast? After reviewing the evolution of the Soviet Union’s strategic posture they concluded that it was “more plausible that the Soviet posture evolved as the result of decisions taken within a large bureaucratic structure than as the output of a small set of individuals working in a highly consistent manner.”77

  The clear implication was that RAND’s efforts to forecast the evolution of Soviet nuclear forces needed to incorporate the effects of actual Soviet organizational behavior—how they really functioned, as opposed to what abstract theory posited. Marshall and Loftus decided to try to build on this insight to improve RAND’s forecasts. In late 1956 or early 1957 they initiated an internal consulting effort. Loftus coined the endeavor Project SOVOY, an acronym he drew from the Russian for “Soviet military” (Советские воиска). The project’s speci
fic objective was to help RAND researchers improve their forecasts of Soviet nuclear forces.

  In the end, SOVOY was less a project than a series of “help memos” to RAND researchers. Nor did the effort prove as successful as Loftus and Marshall had hoped. For many analysts, then and later, the simplicity and attractiveness of the rational-actor model offered a relatively quick and simple analytic path toward a conclusion. Embracing the advice proffered by Marshall and Loftus would not make their work easier, but more difficult. To these objections Marshall would have no doubt responded by noting that estimating military power was not easy, but hard.

  Despite the generally poor reception to their efforts, SOVOY did convince Marshall, Loftus, and some others to begin developing an alternative view of Soviet decision-making, taking explicitly into account the bureaucratic behavior of the various Soviet organizations involved in developing the USSR’s nuclear forces. Whose decisions RAND and US intelligence analysts wanted to predict ranged from those of the Politburo and the Soviet General Staff to the USSR’s military services and their weapons design and production bureaus.* From SOVOY an enduring insight emerged: all of these Soviet organizations, as well as key bureaucrats within them, were competing to have their views and agendas prevail. The notion that one could explain (let alone predict) the behavior of the USSR as the sole product of rational action by a single entity was a mirage, and a potentially dangerous one at that.

  Marshall’s interest in organizational behavior persisted long after his collaboration with Loftus. During the early 1960s he would begin to seek out academics whose scholarship focused on understanding decision-making in large organizations such as major American business firms. He found the work being done by Richard Cyert, James March, and Herbert Simon, along with a small group of scholars at the Harvard Business School, including Joseph Bower, particularly valuable. Their findings reinforced Marshall’s conviction that studying how Soviet organizations made decisions, including the resource shortages and other constraints that affected their force posture choices, was the most fruitful way to begin trying to forecast the USSR’s likely force posture eight or ten years in the future.

  By 1956 Loftus and Marshall were both consultants to the CIA. This not only expanded the sources of intelligence available to both men, but also enabled them to serve as a conduit for some CIA reports that otherwise would not have made it to RAND. In the decade ahead, Marshall’s continuing involvement with the CIA on intelligence issues would bring him other assignments involving nuclear strategy and sensitive intelligence. One of the earliest was the commission headed by H. Rowan Gaither Jr., a San Francisco attorney who in 1957 also sat on the boards of the Ford Foundation and the RAND Corporation. Gaither had been one of RAND’s founding fathers as well as the first chairman of its board.

  Known as the Gaither Commission, the panel got under way in April 1957 when Eisenhower established a Security Resources Panel under his Scientific Advisory Committee. The panel’s assignment was to assess “the relative value of various active and passive measures to protect the [US] civil population in the case of nuclear attack”; in addition, it was asked to study “the deterrent value” of US retaliatory nuclear forces and “the economic and political consequences of any significant shift of emphasis or direction in defense programs.”78

  Marshall was tapped to serve on the Gaither Commission’s staff. Starting in August 1957, he spent five months in Washington working on the Security Resources Panel’s analysis group headed by Robert Prim of Bell Telephone Laboratories and Stanley Lawwill from the Strategic Air Command. Marshall also became involved in an intelligence subgroup that looked at strategic reconnaissance issues. In addition to Marshall, the subgroup included Spurgeon Keeny, who had succeeded Loftus in Air Force intelligence, and Jim Perkins from the Carnegie Corporation. The subgroup’s aim was to explore ways of masking the highly classified or “black” Corona spy satellite program then being developed with “white world,” or unclassified, information.

  The Gaither panel submitted its report, “Deterrence and Survival in the Nuclear Age,” to Eisenhower in November 1957. Scholars have debated how influential it was, especially in affecting US nuclear strategy. What is clear in hindsight, though, is that the report considerably overstated Soviet capabilities. It also portrayed the Soviet economy as expanding much more rapidly than that of the United States, having grown from one-third to one-half the size of US gross national product.

  Working on the Gaither Commission heightened Marshall’s awareness of the fundamental importance of developing an accurate assessment of the Soviet economy’s size, and the portion of it devoted to military programs. Marshall also developed a healthy skepticism as to the way in which these estimates were derived. Over time this skepticism would lead Marshall to engage in a long-running dispute with the CIA over the matter, which he viewed as of fundamental importance to understanding the United States’ competitive position relative to the Soviet Union.

  The main authors of the Gaither report were Colonel George Lincoln, head of West Point’s Social Sciences Department, who had served on Eisenhower’s Project Solarium strategic exercise in 1953, and Paul Nitze, who had also been the principal author of President Truman’s NSC 68, a top secret policy paper outlining US policy for the Cold War. The commission’s report came only a month after the Soviets succeeded in launching Sputnik, the world’s first man-made earth-orbiting satellite. This remarkable scientific feat brought home even to average Americans the vulnerability of their country to nuclear attack: if the USSR could boost a satellite into space, it had the means to launch a nuclear warhead at any location in the United States. News of Sputnik spawned demands in Congress to meet the challenge by, among other things, augmenting US civil defenses, boosting the military budget, and promoting greater education in mathematics and the hard sciences in public schools.

  Moreover, not being privy—as Marshall was—to the intelligence being provided by US U-2 covert reconnaissance flights over the Soviet Union, the Gaither Commission had no way of knowing how far behind the United States the USSR was in terms of intercontinental nuclear forces. In the wake of Sputnik’s success, the Gaither commissioners assumed the worst, portraying the Soviets as having made “spectacular progress” in producing fissile material and jet bombers, and asserting that the USSR had probably surpassed the United States in developing ICBMs.79

  In reality, the first heavy Soviet jet bomber, the Myasishchev M-4, which so worried the commission, proved incapable of flying a two-way mission against the United States. Only 116 of these aircraft were produced by 1960, and most of those were converted to aerial tankers. By comparison, from 1955 to 1962 the United States produced 739 B-52 long-range bombers. As for Soviet ICBM developments, the USSR’s first ICBM, the R-7/7A (designated the SS-6 by NATO), became operational in 1960 or 1961. However, it proved to be an enormous, cumbersome missile. It took twenty hours to prepare for launch and its cryogenic fuel meant that it could not be kept ready to launch for more than a day. No more than six of these missiles were ever deployed.80

  To shore up the US military posture, the commission recommended spending an additional $19.09 billion over 1959–1963 on high-priority military capabilities, plus another $25.13 billion on improvements to active and passive defenses.81 To put these fiscal recommendations in perspective, the proposed five-year total of $44.22 billion exceeded the Defense Department’s total outlays of $41.47 billion in fiscal year 1959.

  Marshall’s work on the Gaither Commission led him to think more deeply about ways of mediating the apparent disconnect between the requirements of nuclear deterrence and those of nuclear war fighting. On the one hand, the more horrific the nuclear devastation US strategic forces could inflict on the USSR, the more likely the Soviets might be deterred from starting a nuclear war. On the other hand, deterrence might fail. If it did, would the United States be doomed? Would it be destroyed as a functioning entity, or could strategic forces be designed and employed to enable outcomes short o
f mutual societal destruction without weakening nuclear deterrence?

  Once back in Santa Monica, Marshall began talking to Goldhamer about this dilemma. They decided to approach it through game-theoretic methods. They constructed payoff, or utility, matrices using largely hypothetical numbers for both the United States and the USSR, the US goal being neither to maximize deterrence nor achieve the best (or least worst) outcome of a nuclear war. Instead they wanted to get the best expected value for both elements combined.82 The values used in the model included the proportion of SAC’s bombers surviving the first Soviet attack, Soviet estimates of SAC’s surviving bombers, US choices of strategy, and both sides’ outcome utilities. To illustrate the methodology, they tested it against a single case—a Soviet surprise counterforce attack against US nuclear forces in CONUS and overseas. Reflecting Marshall and Goldhamer’s interest in organizational behavior, the analysis specifically highlighted differences in US versus Soviet perspectives and strategies. To assist them with the Soviet side of the analysis they brought in Nathan Leites, a RAND colleague who had focused on these issues in a classic analysis of what he termed the “operational code” of the Politburo.

  Some intriguing, if tentative, findings emerged from this effort. Marshall and Goldhamer concluded that if the Soviets attacked first, the best US strategy (and the one the Soviets feared most) would be a mixed targeting strategy in which the United States attacked counterforce targets “using high yield weapons with ground bursts so as to produce extensive fallout” inside the USSR, thereby inflicting heavy civilian casualties as a “bonus.”83 In addition, they found that although tensions between deterrence and nuclear war fighting strategies did exist, they were less numerous, less critical, and more easily addressed than Marshall and Goldhamer had originally supposed.84