Why China Has Not Caught Up Yet: Military-Technological Superiority and the Limits of Imitation, Reverse Engineering, and Cyber Espionage

Abstract Can countries easily imitate the United States' advanced weapon systems and thus erode its military-technological superiority? Scholarship in international relations theory generally assumes that rising states benefit from the “advantage of backwardness.” That is, by free riding on the research and technology of the most advanced countries, less developed states can allegedly close the military-technological gap with their rivals relatively easily and quickly. More recent works maintain that globalization, the emergence of dual-use components, and advances in communications have facilitated this process. This literature is built on shaky theoretical foundations, however, and its claims lack empirical support. In particular, it largely ignores one of the most important changes to have occurred in the realm of weapons development since the second industrial revolution: the exponential increase in the complexity of military technology. This increase in complexity has promoted a change in the system of production that has made the imitation and replication of the performance of state-of-the-art weapon systems harder—so much so as to offset the diffusing effects of globalization and advances in communications. An examination of the British-German naval rivalry (1890–1915) and China's efforts to imitate U.S. stealth fighters supports these findings.

Alexander Gerschenkron. 1 By free riding on the research and technology of the most advanced countries, less developed states can allegedly close the military-technological gap with their rivals relatively easily and quickly. 2 More recent works maintain that globalization, the emergence of dual-use components, and advances in communications (including the opportunity for cyber espionage) have facilitated this process. 3 This literature is built on shaky theoretical foundations, and its claims lack empirical support.
The international relations literature largely ignores one of the most important changes to have occurred in the realm of weapons development since the second industrial revolution : the exponential increase in the complexity of military technology. We argue that this increase in complexity has promoted a change in the system of production that has made the imitation and replication of the performance of state-of-the-art weapon systems harder-so much so as to offset the diffusing effects of globalization and advances in communications. On the one hand, the increase in complexity has signiªcantly raised the entry barriers for the production of advanced weapon systems: countries must now possess an extremely advanced industrial, scientiªc, and technological base in weapons production before they can copy foreign military technology. On the other hand, the knowledge to design, develop, and produce advanced weapon systems is less likely to diffuse, given its increasingly tacit and organizational nature. As a result, the advantage of backwardness has shrunk signiªcantly, and know-how and experience in the production of advanced weapon systems have become an important source of power for those who master them. We employ two case studies to test this argument: Imperial Germany's rapid success in closing the technological gap with the British Dreadnought battleship, despite signiªcant inhibiting factors; and China's struggle to imitate the U.S. F-22/A Raptor jet ªghter, despite several facilitating conditions.
Our research contributes to key theoretical and policy debates. First, the ability to imitate state-of-the-art military hardware plays a central role in theories that seek to explain patterns of internal balancing and the rise and fall of great powers. Yet, the mainstream international relations literature has not investigated this process. 4 Because imitating military technology was relatively easy in the past, scholars and policymakers assume that it also is today, as frequent analogies between Wilhelmine Germany and contemporary China epitomize. 5 In this article, we investigate the conditions under which the imitation of state-of-the-art weapon systems such as attack submarines and combat aircraft is more or less likely to succeed. Second, we develop the ªrst systematic theoretical explanation of why U.S. superiority in military technology remains largely unrivaled almost thirty years after the end of the Cold War, despite globalization and the information and communication technology revolution. Some scholars have argued that developing modern weapon systems has become dramatically more demanding, which in turn has made internal balancing against the United States more difªcult. 6 This literature, however, cannot explain why in the age of globalization and instant communications-with cyber espionage permitting the theft of massive amount of digital data-U.S. know-how in advanced weapon systems has not already diffused to other states. Other contributors to the debate on unipolarity have either pointed to the relative inferiority of Chinese military technology without providing a theoretical explanation, or they have argued that developing the military capabilities to challenge the status quo is, in the long run, a function of political will-an argument that cannot account for the failure of the Soviet Union to cope with U.S. military technology from the late 1970s onward. 7 We argue that in the transition from the second in- dustrial revolution to the information age, the imitation of state-of-the-art military technology has become more difªcult, so much so that today rising powers or even peer competitors cannot easily copy foreign weapon systems. 8 Our ªndings address existing concerns that China's use of cyber espionage and the increasing globalization of arms production will allow Beijing to rapidly close the military-technological gap with the United States. 9 Third, the international relations literature accepts the claim that globalization and advances in communications have made the imitation of military technology easier; yet no one has empirically tested this proposition. 10 This failing is particularly concerning in light of the opportunities opened by cyber espionage-a practice that, according to many observers, could erode the U.S. advantage in military technology. Richard Clark, a former U.S. senior government ofªcial, believes that Chinese cyber espionage could result in the United States "hav[ing] all of [its] research and development stolen"; Gen. Keith Alexander, a former director of the National Security Agency, worries that cyber espionage could lead to "the greatest transfer of wealth in history." 11 With a few notable exceptions, however, international relations scholars have paid little attention to the advantages and limits of cyber espionage for copying foreign military technology. 12 Our research ªlls this gap and tests the conventional wisdom using the case of China, one of the states that has beneªted the most from globalization and that has employed cyber espionage more extensively than any other country. The remainder of the article proceeds as follows. First, we discuss the substantive and theoretical relevance of the outcome we aim to explain: the degree to which rising states are able to close the military-technological gap with their most advanced adversaries. Second, we examine why the increasing complexity of military technology accounts for the variation in outcome we observe over the past 150 years. In the third and fourth sections, we detail the causal mechanisms underpinning our theory. In the ªfth section, we present the results of our empirical investigation: a comparison between Imperial Germany's imitation of big-gun battleships (1890-1916) and China's imitation of ªfth-generation jet ªghters . In the conclusion, we lay out implications of our ªndings for international relations theory and the future of U.S. primacy.

Military-Technological Superiority
Military-technological superiority is a central theme in both international politics and international relations theory. Yet, the discipline has not studied the conditions under which states can close the military-technological gap with their rivals.

military-technological competition
States compete to develop, ªeld, and maintain the most advanced military platforms possible. 13 When a country develops a new military technology, its competitors will devise countermeasures and counter-innovations to limit, and possibly eliminate, the advantage their enemy derives from its innovation. Counter-innovations such as anti-air defense systems force innovators to further improve the performance of their technology. 14 The history of military innovation is, in the end, the history of innovation, counter-innovation, and further innovation. 15 Although countermeasures and counter-innovations can be very effective, they permit countries only to negate the beneªts an enemy gains from its innovations. 16 When countries seek to remain or become regional or global powers, or when they aim to deploy certain capabilities, however, they have to acquire speciªc military platforms, such as aircraft carriers for long-range power projection, jet ªghters for air superiority, or submarines for sea denial. Some countries lagging behind in these capabilities will try to copy others' military innovations and, ideally, to outperform them. 17 Under the right conditions, imitation will facilitate and accelerate the ability of some states to catch up technologically. 18 First, by free riding on the research and technology of innovators, imitators can save the resources they would otherwise need to invest to develop indigenously state-of-the-art technology. 19 Second, imitators can avoid making the mistakes of the innovators or, worse, embarking on technically unfeasible projects-a strong possibility when dealing with cutting-edge technologies. 20 Third, imitators can use their unused resources to improve existing technology and possibly outperform innovators. 21 In sum, imitators will derive an advantage when imitation is cheaper and faster than innovation. 22 These conditions were present in the late nineteenth and early twentieth centuries, when rising powers such as Wilhelmine Germany, Imperial Japan,  23 Implicitly or explicitly, these and similar cases have informed the literature on the balance of power and on the rise and fall of great powers-a literature that largely has accepted the assumption that copying foreign military technology is relatively easy. 24 literature review The imitation of military technology plays a central but unappreciated role in the literature on international relations theory. 25 Internal balancing, for instance, often entails imitating foreign technology. Yet, international relations scholars have not investigated when and why efforts to imitate foreign weapon systems are successful. 26 Instead, they have assumed that states' intentions or incentives to imitate will ipso facto lead to success. 27 According to Kenneth Waltz, for example, because of competition and socialization dynamics in international politics, the "weapons of the major contenders . . . come to . . . look much the same all over the world." 28 Similarly, Robert Gilpin argues that "there is a historical tendency for the military . . . techniques of the dominant state or empire to be diffused to other states in the system." 29 For P.W. Singer, "The problem for 'ªrst movers' . . . is that they have to pay heavily" when developing military technology. In comparison, imitating countries "can 'free ride' on the early cost, copy what works and focus all their energy and resources solely on improving upon what the ªrst mover does." 30 Interestingly, even some of the scholars who have questioned the liter- ature's conventional wisdom agree with the proposition that the imitation of military hardware is relatively easy. In the words of Emily Goldman, "Hardware is often easy to acquire." 31 Similarly, Michael Horowitz writes that "it is [not] difªcult to copy . . . speciªc technologies." 32 Many scholars further argue that globalization and information technologies have facilitated and accelerated the diffusion of technology in the military realm. 33 Joseph Nye observes that, in the age of globalization, "technology . . . eventually spreads and becomes available to adversaries." 34 Some scholars claim that countries with advanced commercial industrial and technological capabilities can exploit their industrial and scientiªc base to develop state-ofthe-art military technology. 35 Other scholars have stressed that, since the early 1990s, commercial research and development (R&D) has supplanted military R&D as the main driver of innovation. 36 As a result, many advanced technologies are now accessible on the global market at moderate cost, including those required for producing ªrst-class military capabilities. 37 For instance, Goldman and Richard Andres maintain that, in comparison to the industrial era, "revolutionary dual-use technologies, like computer and software capabilities," can be imitated more quickly and more easily, because they "are not capital intensive and do not require a huge industrial capacity to exploit." 38 Horowitz warns that because of increasing synergies with the commercial sector, "military technology . . . could become increasingly 'lootable,'" allowing "militaries [to] quickly reverse engineer systems built by other countries. problems with international relations theory There are three problems with the conventional wisdom in international relations theory regarding the ability of states to copy foreign military technology. First, if imitation is easy, why do states invest in military innovation at all? Why not simply wait for others to develop innovations and then copy them? 40 Second, the assumption that imitation is easy is at odds with the literatures in economic history, economics, history of technology, management, science and technology studies, and sociology, which have sought to explain why some innovators retain a ªrst-mover advantage, what the sources of industrial leadership are, and why very advanced companies sometimes fail to replicate an innovation, even with full access to original blueprints and designs. 41 Third, imitating foreign technology also seems to be difªcult in the commercial realm, where, according to international relations theory, it should be particularly easy. For instance, Google and Microsoft, two of the most advanced companies in the world, have struggled to cope with Apple's smartphone and tablet technology. 42 In the following section, we explain why these trends are even more pronounced in weapons production. 43

Complexity and Military-Technological Superiority
Since the second industrial revolution, the complexity of military technology has increased exponentially. This dramatic increase has changed the nature of innovation and of imitation, making the latter much more difªcult to implement. ties also increase-exponentially. 45 Anticipating, detecting, identifying, understanding, and addressing all possible technical problems when designing, developing, and manufacturing an advanced weapon system pose major challenges. 46 Addressing them without creating new problems is an even greater challenge. 47 More challenging still is the need for weapons producers to design platforms that can incorporate cutting-edge and yet-to-be-developed technologies, and to limit their vulnerability to subtle and effective enemy countermeasures and counter-systems. 48 Three developments help account for the increase in the complexity of military technology since the second industrial revolution. First, the number of components in military platforms has risen dramatically: in the 1930s, a combat aircraft consisted of hundreds of components, a ªgure that surged into the tens of thousands in the 1950s and to 300,000 in the 2010s. 49 As the number of components expands, the number of potential incompatibilities and vulnerabilities increases geometrically. Ensuring the proper functioning and mutual compatibility of all the components and of the whole system thus becomes increasingly difªcult. 50 Second, advancements in electronics, engineering, and material sciences have resulted in the components of major weapon systems becoming dramatically more sophisticated, leading military platforms to become "systems of systems." 51 Integrating large numbers of extremely advanced components, subsystems, and systems poses a daunting challenge. More sophisticated components have extremely low tolerances, which in turn require a degree of accuracy and precision in design, development, and manufacturing that was unthinkable a century ago. 52 For instance, aircraft engines in the 1900s and 1910s were "crude" mechanical devices that self-taught individuals could design, assemble, and install in their own repair shops. 53 In contrast, the production of today's aircraft engines is so technologically demanding that only a handful of producers around the world possess the necessary technical expertise. 54 Consider that in turbofan engines, a "close clearance between  manding environmental and operational conditions, thanks to improvements in all metrics (e.g., speed, altitude ceiling for aircraft, and collapse depth for submarines). 60 These improvements, however, have increased the likelihood of technical problems. 61 The more sophisticated a component is, the more likely minor environmental changes will affect its performance. 62 In addition, as technological advances permit weapon systems to operate in once unfamiliar environmental conditions, designers and engineers are forced to deal with previously unknown physical phenomena. 63 change in the nature of innovation and imitation The increase in complexity and the resulting integration challenges have brought about a change in the very nature of innovation. In the late nineteenth and early twentieth centuries, innovation was primarily the product of conjecture, creativity, ingenuity, and intuition, as was the case with two of the most revolutionary military technologies of the twentieth century: submarines (1900) and aircraft (1903). 64 To move from intuition to military platforms, countries needed industrial facilities and managerial capabilities to supervise large production runs-what Alfred Chandler called "the Visible Hand" of industrial capitalism. 65 Because of the increase in complexity, however, innovation has progressively been the result of scientiªc and engineering research, as well as of accumulated experience in design, development, and manufacturing. 66 Consequently, arms producers working on the technological frontier have had to develop in-house technological knowledge bases, or systems integration capabilities-"the Visible Brain" as Keith Pavitt called it. 67 Consider the evolu-International Security 43:3 152 tion of submarines: from "crude, slow and probably as dangerous to their crews as they were to any potential enemy" in the early twentieth century, they are now so complex that "the only comparison you could draw would be a space shuttle." 68 The increase in complexity has also made imitation more challenging. For imitators to have an advantage vis-à-vis innovators, two conditions are necessary. First, the capabilities required to exploit foreign know-how and experience in the production of weapon systems must be relatively easy to develop or to acquire, so that the imitator can swiftly translate foreign designs and blueprints into a working military platform; that is, there must be relatively low entry barriers. Second, the know-how and experience of the innovating country must diffuse with relative ease and with relative rapidity to would-be imitators. The growth in complexity observed over the past century has made these two conditions increasingly difªcult to meet.
During the second industrial revolution, imitating countries could exploit the know-how and experience of their most advanced peers to develop stateof-the-art military technology. 69 In fact, intuition and conjecture could be transferred from country to country with relative ease. 70 The main challenge these countries faced was to mobilize the necessary capital to launch production and to achieve the necessary economies of scale. 71 For industrialized countries, this challenge was not insurmountable. 72 Consider the naval rise of Imperial Japan. Because of its qualitatively modest domestic shipbuilding industry, in 1905 the Japanese navy mostly deployed British-made warships. 73 Yet, through a policy based on "copy, improve and innovate," 74  their subsequent improvement," 75 by 1912-13, Japan was able to commission two super Dreadnoughts that surpassed their British counterparts in both speed and tonnage. 76 And by the 1920s, "the skills available at both naval and commercial dockyards made Japan capable of turning out a range of warships that in design and construction were equal or superior to those of any navy in the world." 77 Over time, however, the increase in complexity of military technology has made the process of imitation much more difªcult. Economies of scale and exorbitant capital investments still represent major barriers for most countries seeking to enter the defense sector. Yet, simply extracting and investing resources is no longer sufªcient to close the technological gap with the most advanced countries. Lacking the necessary know-how for weapon systems production has, in fact, become a major obstacle for actors trying to imitate foreign technology-wealthy countries included. 78 Japan's experience in the 1980s and 1990s offers a useful comparison to the Imperial era discussed above. As Stephen Brooks notes, despite Japan's then primacy in hightechnology and despite several decades of collaboration on weapons production with the United States, its F-2 ªghter proved to be "a white elephant: no better than the F-16C [Japan built upon and] at least twice as expensive to produce." 79 In the next two sections, we explain why the increase in complexity has steadily eroded the advantages of imitation that some countries once enjoyed.

Absorptive Capacity
The increase in technological complexity over the past 150 years has exponentially raised the requirements to assimilate and imitate foreign military technology, thus canceling the ªrst necessary condition for states to enjoy the advantage of imitation-relatively low entry barriers for the imitation of stateof-the-art weapon systems. To free ride on the R&D of a foreign country, a country must be able "to identify, assimilate, and exploit knowledge from the environment." 80 But as scholarship from other disciplines shows, knowledge and experience are not public goods that can be easily and cheaply appropriated. 81 An imitator must possess an adequate absorptive capacity: material and nonmaterial capabilities such as laboratories, research centers, testing and production facilities, a skilled workforce, and a cumulative technological knowledge base (the stock of knowledge acquired through previous projects). 82 Without such absorptive capacity, the imitator will have to develop an advanced industrial, technological, and scientiªc base before it can copy foreign technologies. In the next two sections, we explain how the increase in complexity has created massive and highly speciªc requirements for those seeking to imitate advanced weapon systems.
from limited to massive absorptive capacity requirements Whereas in the past, the requirements to imitate foreign military technology were limited, today, states need to master, to an unprecedented level, a much broader range of disciplines and activities.
limited capacity. In the aftermath of the second industrial revolution, the absorptive capacity required to imitate cutting-edge technologies was comparatively low for great powers. For industrialized countries, capital investments and sufªcient economies of scale were essentially the only constraints, as entering weapons production required relatively little accumulated knowledge or experience. 83 During this period, "within many economic sectors, the knowledge required for moving out of the technological frontier was rather elementary scientiªc knowledge of a kind that had been available for a long time." 84 Even sectors such as metallurgy, which "placed a premium on basic chemical knowledge . . . drew primarily on elementary science when they drew on science at all." 85 This was also true for emerging ªelds such as avia- tion, which owed "practically nothing to the relatively mature state of the science of ºuid dynamics." 86 In addition, the number of disciplines required in a given ªeld was limited. For example, aviation design well into the late 1930s essentially required knowledge about "efªcient aerodynamic structure and hydraulic controls." 87 Similarly, until World War II, submarine development required only reliable and powerful diesel engines, welding of metals, and efªcient designs. 88 In these circumstances, it is clear why "the period before 1914 . . . was unlikely to produce asymmetrical [revolutions in military affairs] that conferred enduring advantages." Technology moved swiftly across borders, and great powers were able to exploit it with relative ease-to the point that it "was indeed often most convenient and cheapest to make use of another country's research and development." 89 For this reason, Germany in the 1930s could move from "possess[ing] no signiªcant aircraft industry" to being "in the forefront of aviation technology" in just three years. 90 massive capacity. The increase in the complexity of weapon systems has exponentially raised the absorptive capacity requirements to assimilate foreign know-how and experience and to imitate foreign military platforms. 91 The stock of accumulated knowledge has, in fact, expanded to the point of becoming a "burden" for those seeking to assimilate it. 92 The number of disciplines involved has increased dramatically, going well beyond those necessary for weapon systems development, and reaching into new, unexplored ªelds related both to the environmental conditions where the platform is expected to operate and to interactions between human beings and technology (e.g., ergonomics, human physiology, and cognitive sciences). 93  must master all of these disciplines, because the margins for error and imprecision have shrunk enormously. Inªnitesimally small mistakes can have potentially catastrophic effects, given the low tolerances and vulnerability of many key components, as well as the demanding environmental and operational conditions in which military platforms are employed. 94 The evolution of aircraft production illustrates this trend. Originally an empirical ªeld, aircraft design in the 1920s and 1930s started to incorporate scientiªc discoveries from aerodynamics theory and engineering science as wooden strut-and-wire biplanes began giving way to streamlined all-metal monoplanes. 95 After World War II, the advent of rocket engines, radio communications, automatic guidance and control, and high-speed aerodynamics created new challenges. In response, aircraft manufacturers had to broaden and deepen their knowledge base to include ªelds such as weapons design, avionics, and material structures, as well as the training of aircrews, combat tactics, and, most importantly, human physiology and atmospheric sciences. 96 With supersonic speed and subsequent advances, the number and sophistication of disciplines required for aircraft development expanded to the point of being well ahead of scientiªc knowledge and understanding. 97 Work on the SR-71 Blackbird exempliªes these trends. Because of the friction resulting from ºying at three times the speed of sound, the body of the Blackbird was exposed to temperatures above 600°F. To address the resulting problems, Lockheed had to develop "special fuels, structural materials, manufacturing tools and techniques, hydraulic ºuid, fuel-tank sealants, paints, plastic, wiring and connecting plugs, as well as basic aircraft and engine design." 98 With the transition to ºy-by-wire, the absorptive capacity requirements grew by an order of magnitude, as aircraft production expanded to a broad set of highly demanding ªelds such as electronics, computer science, and communications, with "software construction [being] the most difªcult problem in engineering." 99 Moreover, given the nature of these disciplines, the margin for error has continued to shrink: a minor glitch in the software or the exposure of the hardware to unforgiving conditions (e.g., extreme heat, cold, or humidity) can be fatal. 100 With the increase in autonomy in military aviation, the number of disciplines required for weapons production has expanded to "unmanned systems, human factors, psychology, cognitive science, communication, human-computer interaction, computer-supporter work groups and sociology." 101 from generic to speciªc requirements Since the second industrial revolution, the absorptive capacity required to imitate foreign technology has become so speciªc that countries can no longer exploit their civilian industries to catch up technologically in the military realm. This change has further raised the entry barriers for imitating advanced weapon systems.
generic requirements. In the second half of the nineteenth and early twentieth centuries, manufacturing beneªted from unprecedented and possibly unique synergies and economies of scope. 102 The relatively low level of technological complexity imposed fairly loose requirements, permitting the adoption across different industries of the same machine tools, the same industrial processes, and the same know-how. 103 For instance, problems related to automobile production were "not fundamentally different from those which had already been developed for products such as bicycles and sewing machines." 104 As a result, "the skills acquired in producing sewing machines and bicycles greatly facilitated the production of the automobile." 105 With mass production, the opportunities for synergies and economies of scope among different industries expanded even further. 106  World War I could easily enter the business of aircraft and tank production by exploiting their existing industrial facilities and know-how. 107 Within a year of starting to produce aircraft engines, Rolls-Royce was delivering a very reliable and high-performing engine (the "Eagle"). Similarly, during the war, the company adapted its "Silver Ghost" chassis, the same used by King George, into an armored car that proved effective during the British campaign in the Middle Eastern desert. 108 Even during World War II, when the level of complexity of military technology was substantially higher than during World War I, the United States and the Soviet Union were able to convert their civilian manufacturing activities to military production at a pace and to a degree that would be unimaginable today. 109 As Richard Overy summarizes, "Manufacturing technically complex weapons . . . [such as] heavy bombers . . . with the methods used for Cadillacs . . . ultimately proved amenable." 110 speciªc requirements. Opportunities for synergies and economies of scope, however, have diminished dramatically. 111 Weapon systems increasingly rely on extremely advanced technologies, such as data fusion or stealth, that in many cases have no application in the commercial sector. At the same time, they operate under uniquely demanding environmental and operational conditions (e.g., ºying at Mach 2). The resulting large number of subtle and challenging technical problems has led to an exponential increase in the required degree of accuracy. 112 Consequently, imitators can no longer exploit their existing technological and industrial capabilities to assimilate foreign military technology: the absorptive capacity requirements have become progressively more speciªc. 113 By "speciªc," we mean that the laboratories, research centers, testing and production facilities, skilled workforce, and the cumulative technological knowledge base developed for a particular type of production cannot be easily redeployed for assimilating and exploiting foreign know-how and experience in weapon systems. 114 For instance, the design of modern weapon systems requires advanced knowledge of enemies' counter-systems, tactics, and doctrines, as well as of the environmental conditions of operations. 115 When copying foreign technology, countries need this knowledge to translate foreign information into actual designs. 116 Design capabilities are also extremely important when seeking to integrate foreign component technologies into a competitive military platform. As John Alic writes, "Early design choices largely determine ultimate performance and costs." Once a design has been chosen, "no amount of analysis, modiªcation, and reªnement can salvage a deªcient concept." 117 Understandably, commercial enterprises do not possess the necessary design capabilities for military production and cannot develop them overnight. 118 In submarine design, for example, "the most challenging competencies . . . require at least ten years of experience and a Ph.D." 119 Moreover, bringing a concept from prototype to ªnished product involves highly speciªc tasks and procedures that share little with other realms. For instance, to perform tests, interpret and understand the results, and identify and address the problems that will inevitably arise when developing new weapons, arms manufacturers need equipment, laboratories, test facilities, and specialized personnel with both domain-speciªc and product-speciªc expertise. These requirements include radars operating at different frequencies, supersonic wind tunnels, know-how in stealth coating, and test pilots with combat experience. 120 Consequently, defense and civilian industries have come to differ dramatically, to the point that even realms such as defense and commercial shipbuilding or aviation have limited opportunities for synergy. 121 Indeed, even military contractors cannot move easily from one area of weapons production to another, as "systems inte- One could argue that through reverse engineering, industrial espionage, or cyber espionage, an imitating country could skip the design and development stages and manufacture a foreign weapon system using its existing industrial base. This argument ignores a key constraint: the increase in complexity has also made manufacturing processes more speciªc and possibly unique. Because of the requirements that military platforms need to meet, today's production processes must achieve stringent levels of precision that are alien to most industries. 123 For instance, the low observability to radar of stealth aircraft will be compromised if "the heads of [just] three screws [are] not quite tight and extend . . . above the surface by less than an eighth of an inch." 124 In the words of the F-117 Nighthawk's program manager, "In building the stealth ªghter, we had to tightrope walk between extreme care and Swiss-watch perfection to match the low radar observability of our original computerized shape." 125 In turn, developing, updating, and preserving this type of manufacturing skill calls for highly speciªc training, practices, and processes. In the shipbuilding industry, workers across all technical skills require "6-8 years to reach at least 90 percent of optimum productivity." 126 The unique requirements of manufacturing weapon systems go well beyond skills and processes. Take, for example, machine tools. In the early twentieth century, as discussed earlier, disparate ªelds such as the automobile and sewing machine industries used the same machine tools. In contrast, since the end of World War II, the production of military aircraft engines has relied on machine tools with a degree of precision that no commercial company possesses or needs. 127 For its part, the U.S. Navy has developed propellers that dramatically reduce the acoustic signature of its submarines. 128  from the John Walker spy ring, the Soviet Union obtained information about how to manufacture U.S. propellers. 129 Their production, however, required computer-controlled milling machines with a degree of accuracy that Soviet machines could not achieve. 130 In short, the Soviets could not rely on their existing capabilities to exploit the information they had obtained through industrial espionage. 131

Technological Knowledge
The second condition for the advantage of imitation requires that imitating countries acquire, relatively easily and quickly, the technological knowledge of how to design, develop, and/or manufacture a given military platform, so that they can take advantage of the innovator's advances before the platform becomes obsolete. 132 The increase in the complexity of military technology, however, has made technological knowledge increasingly tacit and organizational in nature, which means that it does not diffuse to other countries either easily or quickly.
Regardless of how advanced a country's industrial, scientiªc, and technological base is, the production of new military platforms requires work at the design, development, and manufacturing stages aimed at anticipating, identifying, and addressing inherent idiosyncrasies. These idiosyncrasies stem from the challenges related to the integration of state-of-the-art components, subsystems, and systems, as well as from having the platform operate under previously unexplored environmental conditions. As the director of Lockheed's Skunk Works division, Chief Engineer Clarence "Kelly" Johnson, recalled when discussing the SR-71, "Everything about the aircraft had to be invented. Everything." 133 We argue that the technological knowledge related to the production of advanced military platforms such as the SR-71 does not International Security 43:3 162 129. Ibid., p. 73. 130. Ibid., p. 69. 131. In the end, the Soviets managed to copy the propellers, but only after having acquired two pieces of equipment that had been subjected to an embargo: high-precision "nine-axes milling machinery tools" from the Japanese company Toshiba; and a computer device that could be used with them from a Norwegian defense company, Kongsberg. See  diffuse easily, because such knowledge is embedded in the organizational memory of the defense company that produced it. from codiªed to tacit knowledge Over the past century and a half, weapons production has changed dramatically, as indicated by the increase in the development time of weapon systems from a few months to several years and even decades. 134 As noted previously, in the late nineteenth and early twentieth centuries, innovations were the results of conjecture, creativity, ingenuity, and intuition (and sometimes of just plain luck or accidents). 135 This meant that the knowledge of how to produce a given technology was relatively simple; it could be written down in terms of principles and rules-it was codiªable. Codiªability permitted the spread of knowledge. Because of the growing complexity of weapons systems, however, innovations have become the product of extensive prototyping, testing, experimentation, and reªnement: as a result of this change, knowledge related to a given weapon system has become increasingly less codiªable-it has become tacit. As former Secretary of Defense Ashton Carter and coauthors have noted, "Tacit knowledge is a route for maintaining a technological edge in military systems: what cannot be written down can hardly be stolen." 136 codiªable knowledge. During the immediate aftermath of the second industrial revolution, the knowledge behind most innovations was relatively simple: their logic and functioning were directly observable and understandable. 137 As such, an innovation generally carried within itself the very knowhow related to its production process. 138 The simplicity of such know-how, in turn, allowed for the innovation's codiªcation, and hence promoted its diffusion. This is what happened, for example, in manufacturing.
In the mid-nineteenth century, the implicit knowledge of skilled craftsmen was partially codiªed, allowing semi-skilled workers to perform the manual tasks associated with the emergence of mechanized, standardized manufacturing of interchangeable parts. 139  opment of the assembly line, which relied on the simplicity of the processes involved, and thus "the worker's implicit knowledge had to be made explicit." 140 Implicit knowledge was "gathered and analyzed" to permit the fragmentation of work into a multitude of extremely simple tasks. 141 Beginning in the 1910s, the organizational and technical principles of the assembly line and mass production were codiªed in articles and books. 142 The Ford Motor Company, the pioneer of the assembly line, also contributed to the codiªcation of such knowledge when it decided to "have any part of its commercial, managerial or mechanical practice given full and unrestricted publicity in print." 143 Ford's transparency facilitated the diffusion of mass production processes to other industries. 144 A similar trend occurred with the emergence of modern science in the late nineteenth and early twentieth centuries, as scientists began to understand, explain, systematize, and codify the body of knowledge developed by practitioners and on which many industrial processes were based. 145 As a result of the progress of modern science and of its diffusion, implicit knowledge was translated into explicit knowledge, thus reducing and, in many cases, negating the advantages enjoyed by early innovators. 146 tacit knowledge. Because of the increase in the complexity of military technology, the technological knowledge of how to design, develop, and produce a given weapon system has become increasingly tacit. Tacit knowledge cannot be codiªed. 147 It entails knowledge derived mostly from experience and hence is retained by people and organizations: for this reason, it does not diffuse either easily or quickly. 148  face-to-face interaction." 149 To replicate a given weapon system, an imitator needs direct access to the innovator's tacit knowledge-that is, access to the very people who worked on the system. 150 Otherwise, it will struggle to ªgure out what each part does, the requirements it is intended to meet, how to produce it, and how it is connected to other components-in other words, its design, development, and production know-how. 151 Moreover, disassembling a military platform into smaller components to observe and understand its functioning has become signiªcantly more challenging. Today many weapon systems, such as jet ªghters, "comprise highly integrated subsystems that are extremely difªcult (if not impossible) to decompose into independent modules." 152 The introduction of electronics has reinforced this trend, as the functioning of software is governed not by observable physical laws (in contrast, for example, to aerodynamics), but by the software's internal correctness and its perfect integration with the weapon system's hardware. 153 Three factors explain why design, development, and production know-how have become largely tacit. First, today designers, engineers, managers, and specialized workers face an inªnite number of decisions, each entailing inherent trade-offs: from the choice among alternative designs; to the choice among different components, their materials, and their technical properties; to choices about manufacturing techniques and procedures. 154 Identifying the most appropriate choices and solutions relies heavily on experience, judgment calls, and educated guesses-all of which are, by deªnition, tacit. 155 As Marco Iansiti notes, "An employee with thirty years' experience in manufacturing will prob- ably recognize that one design detail will be easier to manufacture than another, but she or he might not be able to articulate exactly why." 156 In fact, many choices are shaped by a subtle knowledge and understanding of existing capabilities, suppliers, and technical properties (e.g., ease of production or maintenance as well the availability of components); by idiosyncratic cultural, environmental, or operational needs; or by contingent factors within the development process (some problems can be solved without redesigning the whole platform). 157 It follows that some of these choices might not be intuitive or rational. Hence, they might be hard to understand by people who have not directly participated in the process that led to them. 158 Second, identifying appropriate choices and solutions entails extensive experimentation, manipulation, prototyping, testing, and reªnement (including a lot of tinkering) with constant back-and-forth among teams of designers, engineers, managers, and specialized workers. This process allows those involved in the production of a given weapon system to explore different options and potential solutions for the many kinds of problems encountered in such work and thus to understand what works and what does not. 159 Because such knowledge is developed in a disorganized as opposed to systematized way, writing it down in terms of general rules and principles is extremely difªcult, if not impossible. 160 Consequently, the very act of developing this knowledge cannot be separated from the knowledge itself, because it reºects the personal experiences of those involved in each stage of production. 161 Third, even if a country had access to all the blueprints and designs of a given weapon system, many crucial aspects would still be lacking, because "the best efforts to describe complex technologies . . . cannot capture all of the details that engineers and technicians understand." 162 In modern weapon sys- tems, relatively small details, such as the speciªc weight of fasteners or the chemical composition of rubber valves, can create huge operational challenges, possibly compromising the reliability and even survival of the platform. 163 For this reason, just one of the many possible problems that weapons development entails can slow down and possibly even prevent a country from replicating a foreign weapon system, even when it has full access to blueprints and designs. 164 For example, in the 1990s and in 2005, the U.S. Navy failed repeatedly to reproduce its material to refurbish its nuclear warheads-codenamed "Fogbanks"-because there were "few records of the process when the material was made in the 1980s and almost all the staff with expertise on production had retired or left the agency." 165 The Navy eventually succeeded, but only after ten years of extensive work and an expenditure of more than $90 million. 166 From this experience, the U.S. Government Accountability Ofªce concluded that "assumptions such as 'we did it before so we can do it again' are often wrong." 167 from individual to organizational knowledge As a result of the increase in technological complexity, single individuals can no longer master all the knowledge and activities required for weapon development. 168 Such know-how and experience have become the product of the collective effort of designers, engineers, managers, and specialized workers with different backgrounds. As such, know-how and experience diffuse very slowly, because organizations are far less mobile than people. 169 individual knowledge. During the second industrial revolution, lone inventors such as Alexander Bell, Thomas Edison, Nikola Tesla, and George Westinghouse brought scientiªc and technological knowledge forward. 170 They understood and put into practice principles that, in some cases, now Why China Has Not Caught Up Yet 167 seem rather elementary but that until then nobody had been able to identify. 171 By 1933, when the legendary aircraft designer "Kelly" Johnson proved that one of the most advanced civilian aircraft of that time was aerodynamically inefªcient, the chief engineer of Lockheed expressed surprise upon learning "how simple the solution [Johnson provided] really was." 172 Moreover, in many cases, the person who designed a new technology was also responsible for developing it or for specifying the principles for its production. As Alic puts it, "Design and manufacturing could be linked in one person's head." 173 Inventors could also transmit their knowledge to their apprentices, to their students, or even to another country in the event they decided to migrate, as Tesla did. 174 organizational knowledge. Because of the increase in complexity, lone inventors have increasingly lost relevance. 175 "By the eve of the Depression," G. Pascal Zackary writes in his biography of Vannevar Bush, "the lone inventor was fading into myth." 176 Laboratories and research centers were becoming the "locus of invention." Bush himself came to chair one of the most important of these centers during World War II, the Ofªce for Scientiªc Research and Development. 177 Since then, designing, developing, and manufacturing advanced weapon systems has become a collective effort involving hundreds and possibly thousands of highly educated and skilled individuals with backgrounds in different ªelds and with decades of experience. 178 These individuals generally possess knowledge about extremely important but very small details regarding these systems. Even the most talented aircraft designer will not be able to develop the software necessary for today's ªghters or to provide instructions on its development. 179 Consider, again, "Kelly" Johnson, director of Lockheed's Skunk Works division from its inception until 1975. Johnson took a leading role in designing some of the most advanced aircraft ever developed, including the P-38 Lightning, the U-2 Dragon Lady, and the aforemen- continue to be celebrated widely in the aviation community. 180 Yet, when ºyby-wire and stealth emerged in the 1970s, even his extraordinary expertise would be put to the test. Johnson believed that the new stealth aircraft that Skunk Works was developing (the future F-117) would never ºy because of its aerodynamically inefªcient shape. 181 Johnson was right: without ºy-by-wire, "Hopeless Diamond," as the aircraft was nicknamed, "would have been hopeless indeed." 182 Without ºight control software that carried out automatic microsecond ºight adjustments, the F-117 "could not even taxi straight." 183 This software was the product of more than twenty years of hard work by dozens of people on computer-assisted ºight control, which then had to be adapted to the speciªcities of the F-117. 184 Today, know-how about weapon systems production is embedded in the collective memories and experience of defense organizations, which severely inhibits its diffusion. 185 The F-117 was designed by a team of ªfty engineers, and still more engineers were necessary to develop and reªne the end product. 186 That number has increased dramatically, with more than 6,000 engineers working on the F-35 project. 187 With numbers this large, if an engineer or a designer decides to migrate or defect, or even if she is kidnapped, she will at best be able to provide only a fraction of the knowledge needed. 188 One might argue that the emergence over the past sixty years of computerassisted design, engineering, and manufacturing, as well as of ªnite element analysis and computational ºuid dynamics, can compensate for the impossibility for individuals to retain and put to use all the knowledge necessary for the development of modern weapon systems. 189 But as Alic notes, this argument "underestimate[s] the difªculties of mastering the new technologies." 190

Why China Has Not Caught Up Yet 169
In fact, "design decisions will continue to be based on judgment, experience, intuition and rough calculations." 191 The lesson is that computers cannot help with conceptualization. 192 Moreover, given the inªnite number of alternative designs, "there are too many permutations to be explored with even the most powerful computer-based methods [so], designers rely, as they always have, on tacit know-how and experience-based judgment." 193 Computer assistance ultimately "requires experts who understand a particular method's domain of application and can interpret the often ambiguous results of complicated mathematical models." 194 The development of the F/A-18 Hornet jet ªghter is revealing in this respect. 195 Computer assistance "failed to predict" a problem the aircraft experienced "and provided little help in ªnding a solution." 196 To address this problem, the prime contractor had to carry out extensive ground and in-ºight testing. 197 In the end, "designers had no choice but to fall back on informal procedural knowledge, some of it tacit." 198 Available evidence suggests that the same limitations apply to artiªcial intelligence and machine learning, as they cannot substitute for human creativity and intuition, but only complement them. 199

Empirical Investigation
In this section, we employ two cases to test our theory. First, we explain the rationale behind our case selection. Second, we examine Imperial Germany's naval rearmament in the context of its naval rivalry with Great Britain from 1890 to 1916. Third, we consider China's aerospace modernization program in the context of its competition with the United States from 1991 to 2018.

research design
The cases of Imperial Germany and contemporary China allow us to put our theory to a hard test and existing international relations theories to an easy International Security 43:3 170 test. In other words, if the conventional wisdom fails here, we should be skeptical of its validity under less favorable circumstances. Conversely, if our empirical investigation supports our argument, we should be conªdent about its success under more favorable conditions. similarities. Imperial Germany and contemporary China represent rising powers with assertive and possibly expansionist geopolitical ambitions. 200 Both countries launched large-scale military buildups with the goal of developing state-of-the-art weapon systems. 201 And in both cases, the rise of commercial enterprises producing dual-use technologies not only drove the overall economic growth of both countries, but also contributed to their military modernization. 202 differences. The two cases exhibit important differences, however, that according to international relations theory should have made China's imitation of U.S. ªfth-generation ªghters signiªcantly easier than Imperial Germany's imitation of the Dreadnought, Britain's all-big-gun battleship. First, although the rise of both countries took place during two unrivaled eras of globalization, China has been operating in a much more open world economy from which it could derive higher inºows of foreign direct investment (FDI) and unrivaled access to foreign technology. 203 Second, Imperial Germany's impressive 88 percent economic growth from 1890 to 1916 pales in comparison to China's 3,092 percent economic growth from 1991 to 2017. 204 205 Third, the German naval buildup relied on the support of the domestic coalition of "Iron and Rye." Once Dreadnought was launched and the Anglo-German naval arms race rapidly accelerated, the German navy had to compromise on battleship costs and capabilities to preserve domestic support. 206 Conversely, China has so far faced little internal opposition to its military modernization program. 207 Fourth, as a land power, Germany could not invest all its resources in naval modernization. 208 Indeed, as Nicholas Wolz notes, nothing "illustrates better the insigniªcance of the [German] naval forces than the fact that until 1888, supreme command of the [German navy] . . . lay in the hands of army ofªcers." 209 As war with France and Russia became more likely, naval budgets were cut. 210 Conversely, since the collapse of the Soviet Union, China has not faced a serious land threat. 211 Thus, the government in Beijing could focus its investments primarily on the hightechnology modernization of both its air force and navy. 212 independent variables. Most important, the cases of Imperial Germany and China display two key differences in our variables of interest: absorptive capacity and access to foreign technological knowledge. When Germany and China launched their military buildups, each had relatively low absorptive capacity: Germany was lagging behind in pre-Dreadnought battleships, and China could produce only old Soviet aircraft designs. 213 Yet, in comparison to Germany, China beneªted extensively from FDI, R&D joint ventures, and International Security 43:3 172 mergers and acquisitions with Western companies, which signiªcantly enhanced its absorptive capacity while allowing it to launch its stealth aircraft program. Moreover, unlike the German case, China has had massive access to foreign technology and know-how, especially given its extensive reliance on cyber and industrial espionage and opportunities to reverse engineer technologies purchased from abroad. 214 In other words, China represents a particularly easy case for testing the argument that globalization and cyber espionage have facilitated the imitation of foreign military technology. 215 The empirical record, however, demonstrates that copying state-of-the-art weapon systems has become only more difªcult. In the span of just three to ªve years, Imperial Germany succeeded in copying the most advanced battleship of its time-the Dreadnought-the product of experience and know-how accumulated by Britain over the previous ªve decades. 216 In contrast, China faced massive struggles in imitating U.S. ªfth-generation jet ªghters: twenty years after its launch, China's stealth aircraft still suffers from shortfalls that make it inferior to its U.S. counterpart. 217 wilhelmine germany and the all-big-gun battleship,  In 1906, at the apex of the Anglo-German naval rivalry, the Royal Navy commissioned Dreadnought, an all-big-gun battleship that delivered higher speed, more stability, better armor protection, and, possibly more important, the ªrepower of two to three pre-Dreadnought battleships. 218 As such, Dreadnought made all existing battleships suddenly obsolete, thus canceling the capital investments that Britain's competitors had made during the past decade, including those of Imperial Germany. 219 Dreadnought battleships incorporated newly available component technologies and translated into practice British Adm. Jackie Fisher's maxim to "hit ªrst, hit hard, and keep hitting." 220 First, Dreadnought battleships employed turbines rather than boilers: their lower weight, increased horsepower, and superior fuel efªciency resulted in greater speed (from 12 knots to a high of around 20 knots) without compromising range. 221 Second, because of developments in mechanics, electrical systems, and metallurgy, Dreadnought battleships carried longer-range quickªring guns that, thanks to the ªrst ªre-control systems, extended the range of naval combat from a few thousand yards in 1905 to 12,000-20,000 yards in 1916. 222 Third, progress in metallurgy enabled the production of lighter but more resistant armor that, combined with developments in naval science, led to a doubling of battleship tonnage (from 15,000 to 30,000 tons). 223 absorptive capacity. For Germany, the challenge appeared daunting; now it had to confront "by far the strongest shipbuilding industry in the world" with unrivaled experience in warship production. 224 In comparison, Germany's absorptive capacity in naval shipbuilding was relatively low in 1906. Until 1876, Germany was still purchasing most of its warships from Britain, and well into the 1880s, it still had no signiªcant defense shipbuilding capability. 225 With the launch, in 1890, of its imperialist foreign policy (Weltpolitik), and the resulting increase in budget allocations to the navy, Germany began to learn important lessons both in weapons procurement and in warship design. 226 Yet, on the eve of the Dreadnought revolution, Germany had only slightly more than a decade of experience in warship production, and its industry was still struggling to produce advanced battleships, which were signiªcantly inferior to their British counterparts. 227 Second, Germany had a late start in reproducing Britain's Dreadnought design and several component technologies. This is because the very factors driving Germany's naval rise-domestic politics and bureaucratic ambitionsparadoxically were also responsible for slowing down, and possibly even harming, its naval modernization. 228 On the one hand, civil-military relations International Security 43:3 174 were unstable: Kaiser Wilhelm II, the great sponsor of German naval rearmament, constantly intervened in navy policy, inadvertently delaying "warship design and construction." 229 On the other hand, Adm. Alfred von Tirpitz, appointed to lead the German Imperial Navy and fulªll the emperor's aspirations, was strenuously opposed to any increase in the capabilities of German battleships, because their additional costs risked undermining the domestic coalition supporting his naval plans. 230 As a result, the German Imperial Navy systematically ignored incoming intelligence about the development of Dreadnought and halted research in the technological domains underlying the new battleship's design, including turbines, ªre-control systems, and longrange guns. 231 For instance, German naval yards continued to develop obsolete designs until late 1905, even though they had been aware since 1903 of the emerging all-big-gun battleship concept. 232 Similarly, until 1910 the German Imperial Navy designed its ªre-control system for short-range naval engagements (around 6,000 yards). 233 Still, on the eve of World War I, no one in the German Imperial Navy "had thought it possible to ªght effectively at a range of over . . . 16,250 yards." 234 As a result, German investments in ªre-control equipment initially remained limited, and the device Germany adopted in 1897 to measure bearing and bearing rate (the StandGerät) was much "clumsier" than its British counterpart (the Dumaresq). 235 Analogous considerations applied to German naval ªrepower: Tirpitz had long preferred "smaller-caliber guns ªring relatively lightweight high-velocity shells." These shells, however, were ineffective at longer ranges, because they "lost velocity more quickly than heavier ones" and thus were unªt for the engagements that took place during World War I. 236 Finally, and paradoxically, Tirpitz's defense industrial strategy explicitly obstructed domestic research on turbines, thus failing to promote the development of an indigenous industry in this domain. 237 access to foreign technology and know-how. The German Imperial Navy had relatively limited access to British Dreadnought-related technologies and technological knowledge-and, when it did, it did not use them. 238 Even though German naval yards had previous access to the all-big-gun battleship design, they did not immediately employ this information to improve either the warship designs they were working on or their design capabilities. 239 In the case of turbines, after having obstructed research in this realm, Tirpitz realized that German industry was lagging. Although he managed to have a British company establish a turbine production plant on German soil, he did not promote the transfer of technological know-how to German companies. 240 outcome. In line with our theory, within a few years, Germany was able to pose a serious challenge to Britain's naval dominance. 241 It made remarkable improvements in ship design and construction as well as in other key component technologies and, in some realms, it even managed to outperform the new British all-big-gun battleships. 242 First, the German turbine industry caught up quickly, enabling the navy's battleships to match Britain's in terms of speed: by 1910-11, even the German companies that did not cooperate with foreign partners succeeded in indigenously producing modern turbines, which, in some cases, proved superior to their British counterparts. 243 Second, with respect to light-but-resistant armor steel plates for ªre protection, German technology outperformed British technology, as the thinner, and hence lighter, 6-inch armor plating produced by Krupp foundries in Germany "was approximately as effective as" the 8-inch steel armor plating manufactured by the Harvey United Steel Company in Britain. 244  technology also caught up quickly: the transition to larger calibers proved unproblematic, and according to some accounts, by the start of World War I Germany possessed the most advanced naval guns in the world. 245 Behind Germany's success was its capacity to transfer Krupp's know-how in steel plating to warship building and to exploit the superiority of its chemical industry for the development of more advanced propellant charges. In fact, Germany's shells enjoyed "greater penetrative power" than the Royal Navy's. 246 Third, Germany succeeded in leveraging its domestic civilian industry to catch up quickly in the realm of ªre-control equipment, which it had initially neglected. 247 After the introduction of the British Dreadnought, the German navy recognized the inadequacy of its own long-range gunnery capabilities, and therefore started working on a rangeªnder (the EU/SV-Anzeiger) more advanced than the StandGerät, which it had acquired in 1897. 248 To assess its new system, the navy acquired a rangeªnder from a British company that, by mid-1907, "had recaptured the lead in ªre-control instruments," and whose equipment (the Mark II) was considered the best available at the time. 249 The tests revealed similar performances between the British and German systems. 250 From 1908 until World War I, the Royal Navy continued upgrading its ªre-control systems (introducing Mark III and Mark IV equipment), but Germany managed to keep pace. By the eve of the 1915 Battle of Dogger Bank, not only was the German ªre-control system "similar in principle" to those of British warships, 251 but "at 20,000 yards," it was also "as accurate as the [British] 9-foot coincidence rangeªnders at 15,000-16,000 yards," 252 while its "spotting procedures could straddle more quickly." 253 In other words, Germany managed to get a lead in ªre-control systems. 254 Combat performance during the most important naval engagement of World War I, the 1916 Battle of Jutland, further supports our conclusions. 255

Why China Has Not Caught Up Yet 177
Despite some material and operational disadvantages that affected the mobility of German warships, 256 the two ºeets performed very similarly in terms of accuracy: "117 British hits (2.45 percent) versus 83 German hits (2.3 percent)," thus suggesting that in terms of overall technology the two ºeets were generally on par. 257 Interestingly, "German ªre started excellently, but got worse as the day wore on, whereas British gunnery improved over time." 258 This indicates that the two sides enjoyed similar technology, but that over the duration of the battle, the Royal Navy could leverage its superior drilling, tactics, and experience.
In sum, despite all the countervailing pressures discussed earlier, 259 Germany managed in just a few years to imitate British Dreadnought battleships and to develop "superior North Sea ªghting ships to their British contemporaries." 260 German ships had in fact "better range-ªnding equipment, superior ªre-control, and improved compartmentalization . . . and better shells." 261 Moreover, German equipment "functioned excellently" during the war, 262 because some of its key components were "apparently better and more plentiful than in Great Britain." 263 china and ªfth-generation ªghter aircraft, 1991-2018 The United States commissioned its ªfth-generation stealth ªghter, the F-22/A Raptor, in 2005. Like Dreadnought, the F-22 had no match when it was ªelded, speciªcally because of its "ªrst-look, ªrst-shoot, ªrst-kill" capabilities. 264 To begin, the application of stealth technology reduces by several orders of magnitude the observability of the F-22 to enemy sensors. 265 In addition, with its supercruising thrust-vectoring engines, the F-22 can achieve fuel-efªcient supersonic speed and high maneuverability, enhancing its perfor-International Security 43:3 178 mance against enemy ªghters. 266 Finally, the F-22 possesses superior and longer-range situational awareness thanks to its advanced onboard computer systems, software, and data-fusion capabilities: these attributes enable the collection, rapid processing, and exploitation of large amounts of different types of data that ultimately increase the F-22's battleªeld performance. 267 absorptive capacity. When in the late 1990s China started work to develop a ªfth-generation ªghter, its absorptive capacity in the military aviation domain was relatively limited. 268 China could assemble modern foreign weapon systems, but its experience with combat aircraft production was restricted to Soviet designs from the 1950s and 1960s. 269 The reasons for China's delay in combat aircraft production include economic backwardness, a command economy, and its international isolation, as well as some particular political choices, such as limiting the amount of attention paid to military aviation during the Cold War, especially when compared to missiles or warships. 270 China did not start from scratch, however. During the 1970s and 1980s, the Chinese defense industry worked on several programs, accruing some experience in military aviation. 271 Moreover, since the 1970s and in particular since the 1990s, China's industrial base has beneªted tremendously from FDI, R&D joint ventures, and mergers and acquisitions with Western aerospace companies. 272 Most of the world's major jet engine producers, airline manufacturers, and companies specializing in avionics and aerospace component technologies have established a presence in China. 273 Chinese companies have also purchased machinery to "manufacture sophisticated weapon systems and related components." 274 Why China Has Not Caught Up Yet 179 access to foreign technological knowledge. Unlike Imperial Germany and its efforts to imitate the Dreadnought battleship, China has beneªted from massive access to foreign technological knowledge in its attempt to imitate U.S. advanced jet ªghters.
China has engaged more extensively in cyber espionage than any other country. 275 In 2007, 2009, and 2011, Chinese hackers entered the servers of the Pentagon and gained access to some ªfty terabytes of data containing the designs and blueprints of U.S. stealth ªghters, as well as other critical information. 276 China has also relied extensively on traditional industrial espionage, including the recruitment of former engineers and scientists who worked for Western aerospace organizations. 277 Together, industrial and cyber espionage have given China extensive access to American know-how. Moreover, China managed to obtain an F-117 that crashed in Serbia in 1999, allowing it to inspect, analyze, and possibly reverse engineer the aircraft's stealth features. 278 Further, since the 1960s, China has beneªted from signiªcant transfers of technology from more advanced countries. 279 In the 1970s, it had access to turbofan engines developed by Rolls-Royce. 280 In the 1980s, Israel provided extensive "weapon-making know-how" useful in the design, ªre control, avionics, and radar capabilities of China's fourth-generation aircraft. 281 Through Pakistan, China gained access to U.S. F-16 Fighting Falcon jet ªghters. 282 After the end of the Cold War, China signed an agreement with Pakistan International Security 43:3 180 Third, military experts agree that China is also lagging in another key realm of ªfth-generation ªghters-avionics. 310 China's difªculties in this realm stem from the fact that aerospace sensors and software development currently pose some of the most daunting engineering challenges. 311 Software problems, in fact, are very difªcult to anticipate: testing and reªning must continue until the software is perfected, given that when the aerospace software fails, it fails "catastrophically." 312 Because China has thus far been unable to copy U.S. aircraft design and engines, there is little reason to believe that it has been more successful in this much more challenging realm. An additional consideration supports this assessment. The complexity of modern aerospace software is unprecedented. 313 Flight control software has become exponentially more complex, eliminating trade-offs in design-for example, between observability to radar and aerodynamic efªciency. In other words, complexity has moved from hardware to software. 314 In addition, software has taken over an increasing number of more complex functions-most prominently, automatic long-range enemy detection, geolocation, high-conªdence identiªcation, and accurate target tracking. 315 Little information is available on Chinese radar operation and data-fusion capabilities, but we can use available evidence on Chinese ºight control as a proxy. This proxy is valid, because developing ºight control software is far less demanding than software for radar operation and data fusion. 316 According to Chinese media, the J-20 still faces problems with its ºight control software. There is thus little reason to believe that China has been able to develop the most challenging part of the J-20s' onboard software. 317 Fourth, the difªculties that China has encountered with the development of less advanced fourth-generation ªghters further corroborate our argument. Despite the signiªcant transfer of technology and support that China has received from Russia, Israel, and various European countries, its industry has struggled in this domain. 318 For instance, in 2004 China suddenly broke a International Security 43:3 184 plane." 328 These estimates show that China has derived a cost advantage of just 14 to 20 percent, which is hardly impressive given that China is the country that has relied most extensively on both industrial and cyber espionage, and that has beneªted massively from the transfer of technology through FDI, mergers and acquisitions, and the purchase of foreign components. Such a cost advantage is even less impressive given that the F-22 is now twelve years old, that the J-20 has signiªcant deªciencies, and that its costs will inevitably increase further as China attempts to ªx its problems and improve its performance. The latter point is critical, because "the ªnal 10 percent striving towards maximum perfection costs 40% of the total expenditure on most projects." 329 The same is true with regard to time. The F-22 became operational in 2005-about twenty years after the project started. 330 Launched in the late 1990s and tested in 2010, the J-20 was ofªcially commissioned in the fall of 2017. 331 Still, it is not yet fully operational and remains inferior to the F-22 on several dimensions: in other words, after more than twenty years, China has not yet closed the gap with the United States. 332 A skeptical reader might wonder whether a more advanced country would have accomplished better results than China in imitating the F-22. The hardest case for our theory would be if the imitating country had the same aerospace capabilities as the United States and if it had access to all the technological knowledge related to the F-22. This case exists. In 2011, the U.S. government interrupted production of the F-22. In 2017, the U.S. Air Force commissioned a study to understand how much it would cost to restart production. In other words, the United States wanted to know what it would take to copy its own technology from just six years before. The ªndings are sobering: the same country that created the F-22 would have to spend $10 billion to restart the production of its ªfth-generation ªghter-equivalent to 25 percent of the total procurement cost for 194 aircraft. 333 International Security 43:3 186 In sum, over the past twenty years, China has made impressive accomplishments in modernizing its aerospace industry. Given the extent to which it has beneªted from globalization and cyber espionage, however, the evidence casts serious doubt on the claim that these two factors have brought about a revolutionary transformation that makes the imitation of foreign weapon systems much easier than it used to be. 334

Conclusion
In his seminal article "Command of the Commons," Barry Posen argued that the unrivaled military-technological superiority of the United States gives it a key advantage over other countries. 335 According to the literature in international relations theory, however, such advantages should be temporary, especially in an era of globalization, real-time communications, and dual-use technologies: knowledge allegedly diffuses quickly, thus undermining one of the sources of U.S. hegemony-superiority in military technology. In this article, we have provided a theoretical explanation for why, three decades after the fall of the Berlin Wall and more than ªfteen years after the publication of Posen's article, U.S. weapon systems largely remain unrivaled.
We have argued that the dramatic increase in the complexity of military technology observed over the twentieth century has signiªcantly shrunk the advantage of backwardness described by Gerschenkron. On the one hand, the requirements for imitating modern weapon systems have become harder to meet. On the other, the technological knowledge of how to design, develop, and produce modern weapon systems has become less likely to diffuse. As a result, compared to the pre-World War I period, today imitating foreign weapon systems is more difªcult. Countries cannot simply free ride on the research and development of the most advanced states: they ªrst have to develop the industrial, scientiªc, and technological capabilities required for becoming ªrst-tier weapons manufacturers; then, they must go through extensive trial and error to address the multitude of extremely small but challenging problems that weapons development entails.
The evidence we presented shows that in comparison to the early twentieth century, when Germany could quickly catch up with Great Britain in all-biggun battleships, in recent years China has faced enormous hurdles in closing