The Role of Mandatory Evacuations as Costly Signals during Interstate Disputes

Abstract International Relations (IR) scholars, particularly those working in the rationalist tradition, argue that costly signalling is one of the main tools that policymakers have to resolve interstate bargaining disputes and, ultimately, to minimize the occurrence of war. Recent rationalist work has greatly advanced our understanding of how costly signalling works in global politics, particularly by unpacking how militarized escalations can signal potential antagonists (e.g. Slantchev 2011). But the current literature is too hasty in dismissing the importance of non-militarized signalling during international crises, particularly for leaders worried about the risk of accidental wars. This paper presents mandatory evacuations (MEs) as a form of non-militarized escalation that states have been increasingly using since World War II to credibly signal their opponents. We illustrate our claims with a case study of China’s preparations for the 1979 Sino-Vietnamese War, when it ordered a massive evacuation along its northern border as a costly signal towards the Soviets.


On the Difficulty of Calibrating Signals in Global Politics
once described the difficult choices state leaders must make during international disputes as a "Goldilocks problem": Nations face the same problem as Goldilocks blundering into the house of the Three Bears. In dealing with problems posed by other states, they can find a response that commits too much, too little, or just the right amount.
The difficulty of calibrating one's responses to those of a potential adversary continues to be an active subject of research by IR scholars, particularly by rationalists working within the framework of the bargaining model of war, as first popularized by Thomas Schelling (1960). In its essence, the bargaining model of war proposes that international crises are best understood as a negotiation between two states, in which each sends offers and receives counter-offers. If a given proposal is accepted by both sides, they implement the agreement; but if one of the states believes it deserves more than the other state is willing to give, it will take its chances through warfare.
In such a situation, both parties have an incentive to dissemble about their goals and to exaggerate their capabilities-in other words, to bluff. This propensity towards "cheap talk" makes it difficult for states to know if the deal they are being offered is a reasonable one or not. One of the few ways out of this dilemma is for states to send each other "costly signals" during crisis bargaining. A costly signal is an attempt to convey information to another party by incurring a cost in delivering the message, and IR rationalists have long held that costly signals are one of the most effective instruments in policymakers' toolkits during international crises. Indeed, Arena (2012: 1) notes that costly signals have "essentially become the textbook recommendation for preventing war." He goes on to caution, however, that while "costly signalling has become a primary prescription for peace, yet our understanding thereof remains underdeveloped." In this paper, we draw attention to one form of costly signalling that has been almost entirely ignored by IR scholars: mandatory evacuations (MEs). We define a mandatory evacuation as a governmental instrument whereby a state orders its own citizens to evacuate from a portion of its own territory because of the near-term likelihood of military conflict in the area. Our definition distinguishes MEs from several other kinds of massive displacements of civilians, and it is worth noting what we do not consider to be MEs. First, although states frequently order evacuations to deal with natural or man-made disasters ("disaster evacuations"), these are not prompted by the possibility of military conflict and so are not included in our definition. Second, MEs are different from "forced evacuations," which occur when advancing hostile forces compel civilians to flee the areas where they live, without explicit orders to do so from their government. Third, state militaries sometimes engage in "non-combatant evacuations" (NCEs), whereby they help extract their citizens from a dangerous situation in a foreign country (e.g., the evacuation of Western civilians from Rwanda in 1994 shortly after the genocide began). An ME occurs, however, only when a state orders the evacuation of part of its own territory. And lastly, we are interested primarily in the mass evacuation of civilians, not government employees (or their dependents), as when embassies and military bases are evacuated. 1 We show below that policymakers have ordered MEs in dozens of international crises since World War II. However, IR scholars have yet to fully consider the role MEs can play in crisis escalation. We argue that, on the spectrum of escalation options available to state leaders during crises, MEs sit comfortably in the middle. Unlike purely diplomatic statements and threats that are essentially costless, MEs entail considerable costs on the part of the states which order them, and hence are a credible signal that the sender is unwilling to back down. However, unlike militarized forms of escalation, MEs do not increase the risk of accidental wars breaking out. Accordingly, MEs represent a useful way out of Rosecrance's Goldilocks conundrum, one that both policymakers and IR theorists need to pay greater attention to.
In Section II, we present a brief recapitulation of the bargaining model of war, before turning to recent work on the effectiveness of militarized vs. non-militarized escalation during international crises. Some IR rationalists have emphasized how military mobilizations are a particularly effective form of costly signalling, but we argue that their accounts neglect the problem of accidental war. Accordingly, state leaders also need to have non-militarized forms of escalation in their policy toolkits, like MEs.
In Section III, we present MEs as a unique policy instrument, as they are escalatory, costly, credible, but nonmilitarized. We show that there is widespread popular belief in the effectiveness of MEs as a signalling device, and explain the logic of how MEs can signal adversaries during international crises.
In Section IV, we briefly review the historical evolution of MEs since World War II. We show that MEs are increasingly being used in the Global South, and specifically as a form of strategic signalling. We then introduce a case study of the People's Republic of China's decision to evacuate over 300,000 of its civilians from along its border with the Soviet Union in the run-up to its 1979 invasion of Vietnam. We show that Chinese policymakers sought to signal the Soviet leadership that they were prepared to defend themselves if the Soviets intervened on behalf of Vietnam, but did so in a way that took pains to avoid the risk of accidental war.
Finally, we conclude in Section V by listing four ways in which taking MEs seriously could help generate new insights for IR rationalists: by prompting a deeper theoretical and empirical mapping of costly signals; by including MEs in existing databases of interstate crises; by exploring the usage of MEs cross-nationally; and by pushing IR scholars to think more deeply about who actually bears the costs of costly signals.

Deepening Our Understanding of Costly Signals: Militarized vs. Non-Militarized Escalation in the Crisis Bargaining Model
In this section, we briefly review the rationalist literature on crisis bargaining, with an eye towards understanding the main different types of costly signals that have been posited.
Rationalists typically frame war as a puzzling outcome in global politics: they assume that there is always a fixed cost to all parties for going to war, so why are the disputants not able to strike a pre-war bargain and avoid paying the costs of war? Take the canonical hypothetical of a crisis between States A and B, which are disputing a given good (typically a contested territory). According to the early treatment set out in Fearon (1995), during the crisis, A and B attempt to obtain the best outcome they can for themselves by bargaining, i.e., by sending the other state proposals for how to split the disputed good. If any of these offers are accepted by both parties, then A and B implement the newly agreed-upon division of the disputed good. At any point, however, either A or B may choose to escalate the conflict using a variety of measures in order to demonstrate their commitment to the disputed good. Alternatively, either state may also choose to end the negotiations by declaring war, at which point the relative military capabilities of each side (along with other related factors like terrain, public support, military leadership, alliance partners, etc.) determine the outcome of the war, which in turn determines the post-war division of the disputed good.2 Fearon and other rationalists expect that most crises should be peacefully resolvable via bargaining. They acknowledge, however, that a variety of factors may dramatically increase the likelihood of war. For instance, the disputed good may not be easily divisible (e.g. a holy site) or one of the parties may doubt another's actual commitment to the post-bargain outcome (the credible commitment problem). Most relevant to our discussion, however, is the problem of private information, also known as asymmetrical information (Fearon 1995: 390-401;Powell 2002: 7;Smith and Stam 2004: 784-788). Each state has an incentive to bluff. Lying about its resolve, capabilities, or expected war costs could allow A to obtain better terms in a non-war bargain than if it was fully transparent. B, however, realizes this situation, and so regardless of what A says, B is probably going to dismiss its signalling attempts as meaningless "cheap talk." Is there a way out of this dilemma? Rationalists argue that there is, by having states send one another "costly signals" during crisis bargaining. A costly signal is an attempt to convey information to another party by incurring a cost in delivering the message. It is because the sender must pay a cost to send the signal that the receiver is likely to view it as more credible than costless "cheap talk." As noted in the introduction, despite their centrality to the crisis bargaining model, costly signals remain surprisingly understudied compared to other aspects of the standard rationalist account (Fuhrmann and Sescher 2014: 920;Quek 2016: 925). For instance, Powell (2002)'s much cited review article on the crisis bargaining model barely discusses costly signalling, and a recent review by Gartzke et al. (2017) calls for an "empirical turn" which "tak[es] the means of signaling more seriously." Insofar as the literature has examined costly signals, it has been noted that they can vary a great deal in form. Fearon (1995: 396) provides many disparate examples: "building weapons, mobilizing troops, signing alliance treaties, supporting troops in a foreign land, and creating domestic political audience costs that would be paid if the announcement proves false." However, instead of studying these diverse types of costly signals in greater detail, the bulk of the rationalist IR literature has followed Fearon's (1997) subsequent, much simpler distinction between two broad sets of costly signals: those where the sender pays the cost of sending the signal upfront ("sunk-cost signals") and those where the sender defers the cost into the future ("hand-tying signals").
To better understand this distinction between sunk-cost signals and hand-tying signals, consider the case analyzed in Fuhrmann & Sescher (2014) of a nuclear-armed state that seeks to deter aggression against a smaller, client state. The nuclear-armed state could deploy some of its nuclear weapons onto the territory of the threatened ally, as happened several times during the Cold War. This would count as a sunk-cost signal, since the sender must pay the costs of the weapons deployment regardless of whether or not war eventually breaks out. Alternatively, the nuclear-armed state could simply sign a defensive pact with the client state, promising to come to its aid should it be attacked. This would be an example of a hand-tying signal, since there is no upfront cost to the nuclear-armed state, but it will suffer repercussions if the client state is attacked and it does not honor its pledge. There would be a direct cost to the nuclear-armed state's international reputation, and furthermore this violation of "audience costs" would likely entail a significant loss of domestic legitimacy for the leadership of the nuclear-armed state. In a democracy, the leadership might face a vote of no confidence or his/her party might lose the next election; in an autocracy, opponents might feel emboldened to begin rioting or even launch a coup d'état.
This ideal-type distinction between sunk-cost signals and hand-tying signals has come under fire lately, with a growing recognition that in practice most costly signals involve elements of both logics. As Slantchev (2011, 28, emphasis added) notes, it is "difficult to find empirical examples that involve pure sunk costs." Indeed, recent rationalist work has emphasized that the most effective types of costly signals will activate both logics simultaneously, that is to say that they will include both up-front costs as well as down-the-line costs.
The costly signal that rationalists have given pride of place to over the last decade has been military mobilizations (see Quek 2016: 935). Militarized escalation is increasingly seen as a logical and necessary part of how states ought to respond to international crises, however distasteful that might seem to some. A strong proponent of this view is Slantchev (2011, 203), who writes in his conclusion: "Military threats can serve a useful purpose in crises. They may reduce the risk of war relative to purely diplomatic actions." Slantchev's argument for why military preparations are so effective relative to other potential signalling mechanisms is that military escalation has multiple, simultaneous effects: We can distinguish among four distinct but related functions the military instrument can have: communicative, committing, subverting, and preparatory. In its communicative role, it can credibly reveal whether one is committed; in its committing role, it can create a commitment by rearranging one's own incentives to fight; [in] its subverting role, it can undermine the opponent's commitment by rearranging his incentives to fight; and [in] its preparatory role, it is just a prelude to war. (96) Other forms of escalation may have some of these features (usually communicative and committing), but no other has all of them. For Slantchev (2011,48), military escalation nicely brings sunk costs and hand-tying together: "the threat embodied in a physical deployment of armed forces is an act that combines features of sunk costs and incentiverearrangement [i.e., hand-tying]." We are sympathetic to Slantchev's conclusions regarding the advantages of militarized escalation as an effective costly signal, and especially to the formal model from which he derives his findings. However, we question whether military mobilizations are always the perfect tool for policy makers to employ during international crises. Specifically, we argue that Slantchev's model downplays the risk that an accidental, unintended war may arise from a militarized escalation.
It is not controversial to note that calling up and moving armed and battle-ready troops nearer to an antagonist's forces during an international crisis may precipitate a war that the mobilizer would have actually preferred to avoid. And yet, the rationalist tradition in IR has been insouciant about the risk of accidental war since at least Fearon (1994, 580). Slantchev's account continues in this tradition, but takes it a step further. It is true that Slantchev (2011, 30-31) devotes several paragraphs to detailing the many ways in which the U.S. and U.S.S.R. both nearly triggered an accidental war during the Cuban Missile Crisis… but he only does so as part of a broader discussion of how it can be rational for state leaders to "leave something to chance" when escalating an international crisis. Clear acknowledgement that war can break out despite policymakers' intentions, not simply because of it, is absent throughout Slantchev's book. Indeed, Slantchev (2011, 111-116) notes that his model counter-intuitively finds that there is less risk of war when antagonists attempt to signal one another via militarized escalation as opposed to non-militarized forms of escalation. We wonder, however, to what an extent Slantchev's surprising finding is simply an artifact of his model's structure, where all the moves occur in a strict sequential order. If the model allowed for simultaneous moves or moves out of order (as happens in the real world), it seems likely that the probability of an accidental war breaking out would be significantly higher. Accordingly, greater caution may be warranted for states considering troop mobilizations as a costly signal (Rider 2009;Quek 2016: 927); in situations where there is a high risk of accidental war, a strategy of reassurance may be the best way forward (Stein 1991).
In short, during international crises, ideally policymakers would have available to them an instrument that is: 1) escalatory (signals the adversary that the state will not back down on a given issue); 2) costly (serves as a credible signal because it imposes costs on the sender); and 3) non-militarized (lessens the possibility of accidental war breaking out). In the next section, we present once such policy tool: mandatory evacuations.3

Introducing MEs as a Costly Signal: An Overlooked Policy Tool During International Crises
In early September 2017, as tensions between North Korea and the international community flared up, aerial footage emerged of Pyongyang, the North Korean capital. A video, uploaded to YouTube, had been shot from a small propeller plane flying surreptitiously over the city. In the video, the capital seemed deserted, with only a handful of cars moving in the streets and few other signs of human activity. A handful of Western media outlets seized upon the scenes as proof that North Korea had instituted an ME of parts of Pyongyang. The New York Post's headline blared "North Korea's capital is a ghost town amid nuclear testing" (Steinbuch 2017). The British tabloid Daily Express went with the even more inflammatory "World War 3 warning: North Korea capital virtually EMPTY shows shocking footage [sic]" (Kerr 2017). Both newspapers cited an earlier piece in the Russian-language Pravda Report, which claimed that North Korean leader Kim Jong Un had ordered 600,000 residents of Pyongyang to leave as preparation for a war with the United States.
As often seems to be the case these days, it turns out that this was "fake news." The aerial footage had been recorded not in September 2017, but rather in May 2016, by a Singaporean photographer who had the permission of the North Korean authorities (Kok 2017). The seemingly desolate cityscape was apparently just the norm for Pyongyang, and independent researchers failed to turn up any corroborating evidence that Pyongyang had ever been evacuated in whole or in part in recent years (Palma 2017).
What was clear, however, is that many observers felt that if Pyongyang were to be evacuated, it would be compelling evidence that the North Korean regime was preparing for the outbreak of hostilities. Indeed, there is a widely-held view-expressed by journalists, generals, politicians, and ordinary citizens alike-that the ME of civilians from a conflict-prone area can serve as a credible signal that a country is preparing for imminent military action. But despite this widespread belief, IR scholars have been slow to acknowledge the signalling function of an ME.
In this section, we clarify our argument about the signalling function of MEs during interstate disputes.4 We propose that when states are unwilling to capitulate during a dispute, but are sensitive to the risks of accidental war, they may choose to order an ME as a costly signal towards their adversary. Making this claim requires us to demonstrate several theoretical assertions: first, that MEs are costly; second, that MEs have a signalling component; and third, that MEs are less prone to accidental war than other major forms of crisis escalation (military mobilizations, force alerts, war games, etc.).
Are MEs costly for states to order, and hence credible in a world full of cheap talk? Yes. States incur at least three distinct types of costs whenever they order an ME. First, there is the direct cost to the state of announcing and carrying out the policy: information must be transmitted to citizens, ways of transporting them away from the area must be provided, some degree of follow-up is needed to ensure that the ME is being obeyed, etc. The second set of costs are borne by the evacuees themselves via the disruption to their daily lives: the costs of the relocation; the disruption of livelihoods and schooling; the long-term psychological effects of being forced to flee; any property damage that may occur while the evacuees are away; and the increased vulnerability to crime and violence that the evacuees may face while displaced. Although these costs are principally borne by civilians, they engender follow-up costs for the state itself: diminished tax revenues from disrupted livelihoods, increased social strife as a result of population movements, higher healthcare expenditures, lower availability of human capital as a result of disrupted education, etc.
Both of these sets of costs must be paid regardless of whether or not war breaks out, and hence follow the logic of sunk-cost signalling.5 The third set of costs, however, relate to domestic politics and may be incurred either immediately or into the future. Because evacuation is quite burdensome for civilians, whenever a government orders an ME it runs the risk of being blamed by the evacuees, who may withhold future political support. For instance, some civilians on the Indian side of the India-Pakistan border, where small-scale MEs are frequently ordered, blame the government for ordering MEs not out of genuine humanitarian concern for their safety, but rather to rouse up patriotic fervor against Pakistan and thereby win votes (Sharma 2016). Regardless of whether their claims are legitimate or not, there is clearly a political cost to ordering an ME. Indeed, even those civilians who are not in the evacuation area observe the government's decision-making, and if they judge that the government ordered an ME for insufficient reasons, they too may withhold future political support. In this way, ordering an ME has a hand-tying function (in addition to its sunk costs), since it often forces governments to act tough during the rest of the crisis as a way of justifying the ME.
Because of these three sets of costs, we believe that MEs are quite costly to the states that order them, and are not undertaken lightly. But is there a signalling component to MEs? We argue that there is. We begin by noting that MEs are inherently public actions-we are not aware of any instances of MEs being secretly undertaken during an international crisis. Indeed, governments usually have an incentive to make them as visible as possible in order to reach all potentially vulnerable civilians, which also helps make them a particularly salient signal for international observers. We concede, however, that MEs are not ordered by governments solely for their signalling functions. Indeed, a state's decision to order an ME is typically an over-determined phenomenon, with four overlapping possible motivations: 1) humanitarian concern for the lives of civilians; 2) military necessity in order to clear out a potential battlefield; 3) domestic political point-scoring; and 4) costly signalling to adversarial states during crises. While the exact mix of motives may vary from case to case, we argue that states are aware of the signalling function of MEs, as our case study below will show.
Do MEs avoid the potential problems associated with militarized escalations? We believe they do. Ordering an ME is an escalatory move, in the sense that it signals that a state is unwilling to capitulate in the dispute. But MEs are inherently less threatening to other states than either troop mobilizations, force alerts,6 blockades, developing or activating special weapons systems, or increasing defense cooperation with allies. The main triggering point for an accidental war is friction between troops on opposite sides operating in close proximity to one another, which is not an issue for MEs.
Overall in this section we have argued that MEs are a rather unique policy instrument, as they are escalatory, costly, credible, and non-militarized. Unfortunately, they remain quite overlooked by IR scholars-we are aware of only a single work in the rationalist IR literature that acknowledges the signalling potential of MEs (Lieberman 2012: 130). This oversight could mean that IR scholars are not fully grasping the strategic interactions during some international crises.

The Historical Evolution of MEs and the Case of the Sino-Soviet Border in 1979
Despite IR scholars' lack of awareness about them, MEs are increasingly being used as a form of strategic signalling during international crises in the Global South.7 In this section, we highlight two trends related to the usage of MEsthat they are increasingly located in the Global South and that they are increasingly serving a signalling functionbefore illustrating our argument with a case study of the Chinese decision to evacuate over 300,00 civilians from along the Sino-Soviet border in 1979.
Although the first recorded instances of MEs date back to Antiquity, such as Themistocles' order to completely evacuate Athens as the invading Persian army approached in 480 BCE, MEs began taking their modern form during the inter-war period in Europe. Confronted with the new threat of airplanes, which could bypass ground defenses and strike directly at vulnerable civilian populations, policymakers in France, Germany, and the United Kingdom engaged in extensive debates about the feasibility and modalities of ordering MEs. Ultimately, millions of women, men, and especially children would be evacuated from major cities and war zones by the French, German, British, and Japanese 6 Force alerts occur when a nation's military readiness is dramatically increased in anticipation of imminent conflict, such as the DEFCON levels in the United States or the Threat Levels in the United Kingdom. 7 Indeed, it is perhaps because MEs are principally done in the Global South that IR scholars-the majority of whom are based in the Global North-overlook them. States that are unlikely to face a foreign threat on their own territory (either because of their isolated geography, status as a great power, or because they are surrounded by friendly states) are unlikely to have cause to order an ME. The United States, which fits all three of those criteria, has never ordered an ME in the lower 48 states since the end of the American Civil War, although it did order an ME of the Aleutian Islands during WWII. We are grateful to Melissa Willard-Foster for suggesting this point to us. governments during World War II (Torrie 2010;Garon 2016). These MEs were not done for strategic signalling purposes, however, since most were undertaken after the war was already well underway. Rather, they were ordered by their respective governments out of a mix of humanitarian concern and military operational reasons.
The advent of nuclear weapons at the end of WWII and into the Cold War significantly altered the debates around evacuation in the West. The scale of devastation that a single atomic bomb could render, as well as the development of ballistic missiles that could reach their targets with only minutes of forewarning, convinced policymakers that large-scale mass evacuations of urban areas was no longer a feasible civil defense measure. Instead, the focus shifted towards "invacuation," or requiring citizens to shelter in place, either in their homes or in designated fallout shelters (Preston and Kolokitha: 7-9). Citizens were strongly discouraged from attempting to flee the city either before or after a nuclear attack.
Accordingly, one major trend in the usage of MEs since WWII has been a geographic shift from the Global North to the Global South, where international crises involving conventional military threats remain far more common. MEs take time to order and perform, and as such require a time scale that missile threats and nuclear weapons render mostly moot.
The other major trend relating to MEs in recent decades is that they have increasingly been ordered for strategic signalling functions. This is largely the result of increases in state capacity, which in turn are linked to technological changes. There are several prerequisites for a state to be able to order an ME (as opposed to forced evacuations, where civilians flee on their own in the face of invading armies): the state must have access to information technologies capable of near-real-time monitoring of the crisis; the state must have access to communication technologies capable of reaching all or almost all of the civilians in the evacuation zone; and the state (or the civilians themselves) must have access to transportation technologies capable of permitting the civilians to leave the zone in a timely fashion. Only relatively recently have most states in the Global South gained access to these technologies and the state capacity to properly deploy them, particularly in the far-flung border zones where most MEs take place. Once states do have access to these technologies, however, evacuations can cease being merely reactive and instead take on proactive qualities, most notably the strategic signalling functions we discussed above.
A comprehensive dataset of MEs does not exist, so it is difficult to know precisely when the first ME was ordered as a form of strategic signalling, or for that matter whether the overall numbers of MEs has been increasing over the years. On the basis of various reports, however, we know that MEs have been ordered during the following crises: -by Egypt during the so-called War of Attrition with Israel between 1967 and 1970, when Egypt evacuated over 90% of the civilian population living near the Suez Canal (Lieberman 2012: 130); -by Saudi Arabia during its stand-off with Iraq during the run-up to the first Gulf War (Browne 1991 In most of these cases, the reasons why the MEs were ordered included a mix of all four of the possible reasons we outlined above (humanitarian, operational, political, and signalling). But we now turn our attention to a different case: China's decision to evacuate over 300,000 of its citizens living near its northern border with the Soviet Union as part of its preparations to invade Vietnam in 1979.8 We focus on this case for two reasons: firstly, as best we can determine, the 1979 Chinese evacuation is one of, if not the, largest ME in recorded history, and as such worthy of the attention of IR scholars. Secondly, the circumstances surrounding the ME are relatively clear and avoid confounding variables, allowing us to illustrate several of the arguments we made above. For most of the Second Indochina War, the militaries of three Communist nations-the Democratic Republic of Vietnam, the Soviet Union, and the People's Republic of China (PRC)-worked hand-in-hand to overthrow the South Vietnamese government and expel the American military from Southeast Asia. All three nations provided men and materiel to the war effort, which culminated with the fall of Saigon in 1975. However, by late 1978 the Chinese had so completely fallen out with their erstwhile socialist brothers-in-arms that the People's Liberation Army (PLA) was directed to begin preparations for a possible two-front war against both the Vietnamese and the Soviets.
China and the Soviet Union had both supported North Vietnam's efforts to unify all of Vietnam during the 1960s, but in reality the two countries had been growing more and more suspicious of one another since the late 1950s. Matters came to a head during the spring of 1969, when the Soviet Union mobilized twenty-five divisions along the Sino-Soviet border-some equipped with tactical nuclear weapons-who faced off against over 800,000 soldiers of the PLA, which was at the time deep in the throes of the Cultural Revolution (Nathan and Ross 1997: 44). Multiple border clashes occurred throughout 1969, as unpremeditated gunfights between platoons on opposing sides often threatened to spark a larger conflagration, despite the wishes of senior political leaders in both Moscow and Beijing. The Soviet leadership seriously considered using nuclear weapons against China, making it clear to all that the PRC and the Soviet Union now considered each other as strategic threats (An 1973). Both the Chinese and the Soviet militaries remained forwardpositioned along their long border throughout the 1970s; not until Gorbachev's "New Thinking" in the mid-1980s would a coordinated military drawdown begin on both sides (Nathan and Ross 1997: 49).
China's relationship with the newly-unified Socialist Republic of Vietnam also unravelled over the course of the 1970s, as documented in detail in Khoo (2011) andZhang (2015). Despite's the PRC's rhetoric of regarding Vietnam as a "little brother" throughout the Second Indochina War, Vietnam's leadership grew increasingly mistrustful of the giant to its north as the decade wore on. Vietnam instead turned to the Soviet Union for economic and military assistance; it allowed Soviet ships and planes to begin making use of the abandoned American bases in the country, and it signed the Soviet-Vietnamese Treaty of Peace and Friendship on November 3, 1978, which China perceived as a deliberate effort by its two former allies to encircle it (Pao-Min 1985: 68-69;Zhang 2015: 48-49). Other factors also caused Sino-Vietnamese tensions to reach a boiling point in 1978: Vietnam expelled 200,000 ethnic Chinese whom it regarded as a potential security risk, and Chinese officials counted 1,100 border clashes over the course of that year (Khoo 2011: 125;Zhang 2015: 41). Accordingly, Deng Xiaoping-who had recently re-ascended to the uppermost levels of the Chinese Communist Party following his purge during the Cultural Revolution-directed the Chinese Central Military Commission to begin planning for a war to "teach Vietnam a lesson" in late 1978 (Zhang 2015: 40-66). The last straw which convinced the Chinese leadership of the need for a punitive war was Vietnam's invasion of Kampuchea in December 1978; China viewed the Khmer Rouge regime in power in Kampuchea as a friendly client state, and was aggrieved by Vietnam's effort to oust them (Pao-Min 1985: 84).
As China began preparing to invade Vietnam, one of the biggest questions confronting its military planners was how the Soviet Union would react. Would the Soviet Union honour the mutual defense provisions of its recently signed treaty with Vietnam? Would the 54 Soviet divisions positioned along the Sino-Soviet border, which in 1978 had conducted large-scale military exercises simulating an attack on China, turn their war-gaming into reality (Khoo 2011: 105;Zhang 2015: 50-51, 62;Nishihara 1980: 73 reports the figure as 44 divisions)? Would the Soviets attempt to use their nuclear weapons to sway the balance of the war in Vietnam's favor? The uncertainty surrounding the "Soviet Factor" profoundly shaped China's diplomatic and strategic planning for the invasion, in at least three ways (Burton 1979: 713-714;Zhang 2015: 48-51, 64-65).
First, China's leaders sought assiduously to characterize their invasion as a highly circumscribed war, limited in its objectives, duration, and methods. Deng Xiaoping and other top Chinese leaders publicly described the war as a "self-defense counterattack" against Vietnamese "provocations," as well as "a retaliatory operation" which would be "limited in time and space"-all phrases aimed at placating the Soviets and the rest of the Communist world (Chen 1983: 240;Khoo 2011: 133;Zhang 2015: 64). Beijing publicly stated at the outset of the war that its goal was merely to "punish" Vietnam, and while the exact form the chastisement would take was left vague, Chinese leaders explicitly foreswore holding on to "a single square inch of Vietnamese territory" and a week into the war also stated that Chinese forces would not advance into the crucial Red River Valley where Hanoi was located (Jenks 1979: 804;Pao-Min 1985: 91;Zhang 2015: 64). In terms of military tactics, the Chinese never used their aircraft offensively at any point during the war, despite a large edge in numbers; military historians dispute, however, whether this was to emphasize the war's limited nature or for fear of Vietnam's robust anti-air defenses (Donnell 1980: 24;Jacobsen 1981: 101;Zhang 2015: 123-126). China continued giving public assurances about the limited nature of its aims and tactics as the war progressed (Chen 1983: 251;Khoo 2011: 134), eventually declaring that it had met its objectives and beginning a military withdrawal on March 5, only 17 days after the war began.
Second, China's military made serious preparations to deal with any possible Soviet counter-invasion. China's best military divisions were actually not sent south, but rather kept north along the Sino-Soviet border, leaving the task of invading Vietnam to less well-equipped and less well-trained units (Gompert, Binnendijk and Lin 2014: 123). And China chose to keep the bulk of its troops stationed along its northern border rather than its southern one: while perhaps 330,000 Chinese troops massed along the Sino-Vietnamese border (about 10% of China's active duty ground forces), up to 1.5 million troops remained positioned along the Sino-Soviet border (Chen 1983: 243;Scalapino 1986: 28). Even the invasion's timing was chosen with an eye towards the Soviets: the February spring thaw in China's north would likely have mired any invading Soviet divisions in deep mud (Nishihara 1980: 71-72;Zhang 2015: 64). As the start of the invasion drew closer, China's northern and northwestern military regions were ordered to be on maximum alert (Chen 1983: 243;Zhang 2015: 59), and the Chinese received daily intelligence briefings (including satellite imagery) from the United States of Soviet deployments along the border (Khoo 2011: 133;Zhang 2015: 62). Deng Xiaoping summarized his attitude towards the Soviets at a conference of high-level party officials the day before China's attack: "We have long ago made full preparations for a Soviet invasion. If they attack on a big scale, we shall fight them on a big scale; if they attack on a medium scale, we shall fight them on a medium scale; and if they attack on a small scale, we shall fight them on a small scale" (cited in Pao-Min 1985: 88).
That being said, the Chinese leadership absolutely did not want the Soviets to intervene militarily. Accordingly, the third-and for the purposes of this paper, most crucial-set of actions the Chinese war planners undertook was a massive, pre-war ME of large swathes of the Sino-Soviet border, particularly in the provinces of Heilongjiang and Xinjiang. The number of civilians evacuated is usually estimated by various sources at around 300,000,9 which probably qualifies the Chinese evacuation during the winter of 1979 as the largest ME since WWII. Despite its scope, however, the English-language scholarly literature completely ignores the ME, except for passing mentions in accounts of the Sino-Vietnamese War proper.
February 1979 was not the first time that China had ordered an ME of its side of the Sino-Soviet border. Others had taken place at particularly tense moments in Sino-Soviet relations, most notably during the summer of 1969, when "extensive civil defense measures (digging of air raid shelters, partial evacuation of cities) was [sic] initiated in many parts of China" (Hinton 1971: 59). But the timing and scale of the 1979 ME strongly suggest that the ME was a calculated strategic signal intended to convey to the Soviets that China did not want war to break out on its northern border.
Indeed, because of its scale it is difficult to explain the ME in terms of any rival explanations, such as humanitarian concerns, military operational considerations, domestic political reasons, or routine standard operating procedures. The size of the Chinese territory emptied of civilians dwarfed the areas that the Soviet military would have been able to seize (at least in the initial few weeks),10 ruling out an explanation centered on military operational necessity. Similarly, it is difficult to argue that the ME was primarily undertaken for humanitarian reasons, given that the lives of many civilians were disrupted who were otherwise at minimal risk even if a large-scale Soviet invasion had been forthcoming. Nor can we easily conceive of a story where the ME would have been ordered for domestic political reasons, and no historical accounts of the period have yielded such an interpretation. Instead, we believe that the Chinese state deliberately imposed significant costs upon itself and its civilians out of proportion with the expected military calculus of the war precisely in order to facilitate the signalling dimensions of the ME. The senior military and political leadership in Beijing well remembered the spring of 1969 ten years earlier, when they had struggled to prevent an accidental war with the Soviets, and on this occasion chose to accompany their military mobilization with a defensive and non-threatening ME.
Our argument is strengthened by noting that at no point before or after its invasion did China order an ME along its southern border with Vietnam. Indeed, no ME was ordered in the south despite the fact that raiding units of Vietnamese soldiers often crossed into China's territory during the war (Jacobsen 1981: 101;Chen 1983: 253;Khoo 2011: 134). Thus, the Chinese did not order an ME in the south despite a clear military case for doing so, but did order a massive one in the 9 While later academic texts have cited this same figure, the earliest citations for it seem to be Chen (1983): 243 andPao-Min (1985): 89. Chen and Pao-Min themselves cited contemporaneous Western and Asian newspaper articles. 10 Chinese pre-war intelligence analyses suggested that even if the Soviets did decide to massively intervene on Vietnam's side in the war, it would have taken them at least a month to transfer enough divisions from Eastern Europe to overwhelm the Chinese (Zhang 2015: 50-51). north, quite out of proportion with the military situation. The most likely reason for this is that China was very sensitive to the risk of an accidental war breaking out in the north but intended on fighting a war in the south.
Did the Soviets receive the message? It is difficult to know for sure, in part because the bulk of Soviet records that could shed light on how the U.S.S.R. perceived China's actions remain sealed and unavailable to researchers.11 What is clear is that, although the Soviet Union ultimately chose not to intervene militarily during the Sino-Vietnamese War, they did carefully monitor the course of the conflict and sought to ratchet up pressure against the Chinese via a variety of measures (Nishihara 1980: 74-75;Donnell 1980: 25;Jacobsen 1981: 97-102;Chen 1983: 251-252). Gompert et al. (2014: 125) frame the Soviet's calculus in explicitly rationalist language: While the Soviet Union engaged in a series of deterrence signals-which included alerting its troops in the Far East, intensifying surveillance, conducting live fire exercises, sea-and air-lifting supplies to Vietnam, and increasing its naval presence in the South China Sea-it did not open a second front.
In other words, the Sino-Soviet crisis related to the 1979 War was very much one between two rational actors attempting to influence one another via strategic signalling.12 And some analysts see clear evidence that China's signalling prior to and during the war had an impact on the Soviet Union. For instance, Khoo (2011: 134) writes that: Beijing made sure to send out signals to the Soviets that China's use of force would be limited in scope and intended to deliver a punishment to the Vietnamese. Comments by Soviet diplomats strongly suggest that they understood these Chinese signals.
Overall, the current lack of direct archival evidence makes it difficult to arrive at definitive conclusions about the impact of China's 1979 ME. We have sought to show that the ME was ordered not out of humanitarian concern, for military purposes, for domestic political reasons, or as a matter of standard operating procedures, but rather to signal the Soviet Union not to intervene militarily. If China's decision to order an ME along its northern border during its invasion of Vietnam did affect the Soviet decision not to intervene militarily, it would immediately qualify as probably the most effective signalling ME in recorded history. In 1979, the two most powerful countries in Asia came dangerously close to open warfare during a particularly tense moment in the Cold War: if an ME played even a small role in preventing that from happening, IR scholars should take note.

Developing a Better Understanding of MEs in the Academy
We have argued in this paper that MEs are a costly, credible, and non-militarized form of escalation that states can and do use during international crises. In this concluding section, we flesh out the implications of our research for the discipline of IR, and particularly for the rationalist literature in IR. Specifically, we highlight four ways in which taking MEs seriously could help generate new insights and research agendas for IR rationalists.
First, we urge IR rationalists to pay more attention to identifying and cataloguing costly signals. As noted above, despite their importance in rationalist accounts of the crisis bargaining model, costly signals remain severely understudied in IR, and a thorough mapping of them-at both the theoretical and empirical levels-is sorely needed. We have identified MEs as one form of strategic signalling that IR scholars currently overlook, but there may be others. For instance, we have not yet seen a comprehensive account in IR of the strategic signalling aspects of either "force alert levels" or "no-fly zones," despite some work that suggests that policymakers may consider them as types of costly signals (Meibauer 2017). Alternatively, it should be seen whether the general logic of MEs we have sketched out in this article also applies to embassy and military base evacuations. For instance, when the United States announced in March 2019 that it had withdrawn all of its diplomats from Venezuela because of a bitter dispute about the legitimacy of the Venezuelan government, was that perceived as an escalatory signal by observers in Venezuela and elsewhere? We believe this is likely, although because the costs of ordering an embassy evacuation are significantly less for the sender than ordering an ME, the effectiveness of the signalling will be proportionately lesser.
Second, we believe that the core databases that many IR rationalists rely upon to test their theoretical claims about state behavior during crises need to be extended. In particular, the Militarized Interstate Disputes (MIDS) dataset should be updated to include the ordering of MEs as one of the possible measures that disputing states can take in its classification scheme. In general, IR scholars who quantitatively study the strategic interactions of states need to be more aware of MEs as a tool that policymakers use, lest they only end up studying a portion of the overall story.
Third, we invite IR scholars to investigate our claims about mandatory evacuations cross-nationally. Beyond the PRC, are other states also making use of MEs as a costly signal? Are policymakers aware of the signalling possibilities that MEs afford during interstate disputes? Are some types of states/governments/militaries more likely to employ MEs as a form of strategic signalling than others?
Fourth, we urge IR scholars to think in a deeper fashion about how exactly costs are imposed when states seek to send costly signals. The blithe language IR rationalists employ about "credibility," "costs," "signals," and "deterrence" belies the very real suffering borne by real people-the very citizens that states are ostensibly seeking to protect-when states order MEs. The costs to evacuees go beyond the mere monetary costs of relocation to include the disruption of livelihoods and schooling, long-term psychological effects, and heightened vulnerability while away from their homes. Indeed, those frequently displaced by MEs, such as civilians along both sides of the India-Pakistan border, may in some ways suffer more than long-term IDPs: because of the stress of constant relocation, because of the pressure to willingly bear the costs out of patriotism, etc. Ordering an ME may sometimes be the most rational and humane course of action for states in a given circumstance, but IR scholars would do well to treat this intriguing state practice with sober interest.