OUP user menu

Security Commitments and Nuclear Proliferation

Dan Reiter
DOI: http://dx.doi.org/10.1111/fpa.12004 61-80 First published online: 1 January 2014


This article develops a theory connecting security commitments and the decision to acquire nuclear weapons. In a threatening environment, third party security commitments can reduce a state's fear of abandonment in the event of war and its motive for acquiring nuclear weapons. However, a threatened state may reject at least some kinds of security commitments, such as foreign deployed nuclear weapons, if it fears that such commitments increase the risks of entrapment, the possibility that the threatened state will be dragged into a war it would like to avoid. The article looks at three kinds of security commitments, alliances, foreign deployed nuclear weapons, and foreign deployed troops. In quantitative tests, it finds strong evidence that foreign deployed nuclear weapons reduce proliferation motives, only very limited evidence that alliances reduce proliferation motives, and no evidence that foreign deployed troops reduce proliferation motives. It also presents several qualitative evidence, which supports the quantitative evidence, and in particular helps explain why alliance ties sometimes do not prevent proliferation.

Though most scholars recognize that nuclear proliferation is driven at least in part by international insecurity, there has been relatively little work exploring how insecurity affects the decision to acquire nuclear weapons. Qualitative work often challenges the security model of proliferation, and quantitative empirical work tends not to go beyond including simple controls for security.

This article develops our understanding of how international insecurity affects the decision to acquire nuclear weapons. Specifically, it asks: what are the relationships between third party security commitments and nuclear proliferation? Answering this question is of both academic and policy interest. On the academic side, answering this question may explain why some states even in highly threatening international environments elect not to pursue nuclear weapons and why some states reject offers of security commitments. Answering it may also provide a more nuanced understanding of the sources of and antidotes to insecurity.

On the policy side, security commitments including in particular forward deployment of American nuclear weapons as a means of reassuring allies are part of contemporary policy debates on nonproliferation. A 2010 Congressional report related that some American allies located near Russia believe that the forward deployment of American nuclear forces is “essential to prevent nuclear coercion by Moscow…[and maintain] the credibility of the U.S. extended deterrent” (America's Strategic Posture's 2010:20–21). Baltic leaders strongly favor keeping American nuclear weapons in Europe (Benitez 2011; Kyl, Johnson, Sessions and Crapo 2011).

Forward deployment of US nuclear weapons remains an issue in South Korea, where US nuclear weapons were deployed until 1991. Recent acts of North Korean aggression such as 2010 attacks on a South Korean naval vessel and a South Korean island have enhanced South Korean insecurity. Some South Korean politicians have requested the redeployment of American nuclear forces to South Korean territory, and others have suggested South Korea consider acquiring its own nuclear arsenal (Ramstad 2011).

This article develops a theory of the relationships between a state's decisions to accept security commitments and a state's decision to acquire nuclear weapons. Accepting a third party security commitment presents a state with a trade-off of risks. Security commitments can reduce the risks of abandonment, the possibility that an attacked state receives no international assistance in the event of war. However, commitments can also increase the risks of entrapment, the possibility that a security commitment would involve a state in a war it would prefer to avoid. If a threatened state procures security commitments, it reduces the risks of abandonment but increases the risks of entrapment. If a threatened state acquires nuclear weapons, it solves the abandonment problem by providing for its own defense and avoids the entrapment problem. However, acquiring nuclear weapons may impose other financial or diplomatic costs. The basic propositions of the article are that states with high abandonment fears and low entrapment fears are more likely to prefer security commitments over acquiring their own nuclear arsenals, that states with high abandonment fears but without an offer of sufficiently credible or powerful third party commitments are more likely to acquire nuclear weapons, and that states with high abandonment fears and high entrapment fears are more likely to acquire nuclear weapons, even if third party security commitments are offered.

The article contains two modes of empirical analysis. First, it presents quantitative statistical analysis of the causes of nuclear acquisition for all states for the years 1945–2000. Using new data on the foreign deployment of nuclear weapons, it finds that a state is significantly less likely to acquire nuclear weapons if foreign nuclear weapons are deployed on its soil. Specifically, no state has ever acquired nuclear weapons while it hosted foreign deployed nuclear weapons. The quantitative analysis generally casts doubt on the proposition that defense pacts make nuclear acquisition less likely. It finds no evidence that the foreign deployment of troops makes nuclear acquisition less likely. The statistical analysis provides improvements over some past quantitative studies, because of the inclusion of an important independent variable previously neglected (foreign nuclear deployments), because of its exploration of the independent effects of the United States, Chinese, and Soviet alliances, and because of some other small improvements in research design.

The second mode of analysis is qualitative, examining the proliferation decisions of Britain, France, Germany, Pakistan, South Korea, and others. It finds support for several of the hypotheses generated by the theory, including that high abandonment fears coupled with an absence of foreign nuclear commitments can encourage nuclear proliferation. This section also provides more nuanced insight into understanding why alliances with nuclear powers sometimes fail to reassure fearful states considering going nuclear. The qualitative analysis also found support for the hypothesis that states with high abandonment and high entrapments fears might reject foreign nuclear deployments in favor of developing independent nuclear forces.

The article proceeds in four parts. The first section presents the theory. The second section presents the quantitative tests. The third section discusses qualitative evidence. The final section concludes.

Security Commitments and Nuclear Proliferation

Acquiring nuclear weapons can be costly, in terms of financial demands, an increased likelihood of international diplomatic censure and economic sanctions, and enhanced danger of regional arms races. Why then do states go nuclear and pay these costs? Scholars have developed a number of different theories of why states acquire nuclear weapons, focusing on elements such as civil-military relations, domestic politics, globalization, norms, psychological factors, and others (see Sagan 1996/97, 2011; Hymans 2006; Solingen 2007). Perhaps the longest-standing theory of why states acquire nuclear weapons is external insecurity. This theory predicts that as states feel greater nuclear or conventional threats posed by other states, they become more motivated to acquire nuclear weapons. It views the history of nuclear proliferation as a series of reactions to security threats: Germany's World War II nuclear program caused the United States to go nuclear; the American nuclear program caused the Soviet Union, China, and North Korea to go nuclear; the Soviet nuclear program caused Britain, France, and perhaps South Africa to go nuclear; China's nuclear threat caused India to go nuclear; India's threat caused Pakistan to go nuclear. The Arab conventional threat and perhaps the Soviet nuclear threat caused Israel to go nuclear, and the Israeli nuclear program in turn stimulated nuclear programs in Iraq, Libya, Syria, and Iran. This is the fundamental arms race dynamic in international relations, closely related to the security dilemma and the spiral model of war. Absent higher levels of insecurity, states wish to avoid the financial and diplomatic costs of going nuclear (on the insecurity explanation, see Sagan 1996/97; Solingen 2007; Hymans 2006; Kramer 1998).

Third parties can alleviate a state's sense of insecurity by making commitments to that state's defense. Such commitments are sometimes called “extended deterrence” and have comprised a substantial portion of post-1945 American and Soviet foreign policies (Schelling 1966). Extended deterrence can have verbal and/or physical manifestations and frequently has one or both of two goals: prevention of attack on the ally, and reassurance of an ally sufficient to dissuade the ally from going nuclear (on reassurance and nuclear proliferation, see Betts 1977). Security commitments can reduce fears of abandonment that in the event of war, third parties may stand aside and let an attacked state fend for itself.

However, third party security commitments are not without risk, even for threatened states. Security commitments can increase a state's risk of entrapment that some forms of third party security commitment may force the threatened state to become involved in a conflict it otherwise would rather avoid. Security commitments might also mean the preliminary and unwanted escalation of conflict (Snyder 1984). Entrapment is not unrelated to the idea of alliances as “chain-gangs,” that if one alliance member enters war, it may drag other members into war along with it (Christensen and Snyder 1990). Entrapment fears may push a state to reject an offered third party security commitment, if entrapment risks outweigh the benefits of reducing the probability of abandonment.

Consider two forms of contemporary external security guarantees, and their abandonment/entrapment trade-offs. An alliance is a formal commitment for the signatories to aid each other in time of war. At the dawn of the Cold War, for example, the United States built a global network of formal alliances such as NATO to reassure friends, deter Communist aggression, and reduce the incentives of states to acquire nuclear weapons (Kissinger 1957:264). However, some believe that alliances are limited in their abilities to reduce abandonment fears. There is no international court or government that can compel a state to abide by its alliance commitments. Further, states sometime interpret the language of their treaties narrowly to justify staying out of conflict. For example, in the 1950s, France tried to convince the United States that Algeria was part of France, thereby engaging America's North Atlantic Treaty obligation to intervene on France's behalf in its colonial war in Algeria. However, the United States disagreed with this assessment and remained uninvolved (Costigliola 1992:111). Doubts about alliance agreements might be even greater in a nuclear context, as an ally might be less likely to honor an agreement if doing so risked triggering a nuclear war that might engulf the ally as well (Schelling 1966).

Beyond reducing abandonment fears, alliances can also raise entrapment risks. Some states fear that joining an alliance would compel undesired involvement in war involving other signatories and hence reject opportunities to join alliances. Entrapment fears pushed Switzerland, Norway, the Netherlands, and other European states to remain neutral in the 1920s and 1930s. In the post-1945 period, severe entrapment fears caused Sweden to reject an invitation to join NATO, despite heavy American pressure (Reiter 1996).

A second form of security commitments is military deployments. The deployment of military troops on the soil of another nation could serve to reassure that nation. Directly, those troops help defend the nation from attack. Indirectly, those troops would likely trigger the third party's involvement in the event of war. As soon as these troops are involved in combat, and especially as they start to experience casualties, this will incur the third party's involvement, serving as a “trip wire” or “plate glass window.” Some recognized that the relatively small American troop deployments during the Cold War in locations like West Berlin would be unlikely to defeat an invasion force, but could still deter a Soviet attack by ensuring American involvement in the event of war. As Thomas Schelling (1966:47) put it, though those troops cannot stop an invasion force, they do have a function: “Bluntly, they can die.” Notably, the foreign deployment of foreign troops incurs fewer entrapment risks than international alliances. Such deployments generally do not create risks of involvement in wars occurring outside of the host state's territory. That said, the effectiveness of such trip wires has been contested (Giordano 2004).

An additional type of military deployment that can serve as a security commitment is basing nuclear weapons on the territory of the insecure state. Verbal threats to use nuclear weapons in defense of an ally may become more credible if they are coupled with the physical presence of nuclear weapons on the ally's territory (Wohlstetter 1961:378). Though nuclear weapons based off the territory of an ally (such as sea-based weapons, or intermediate range or intercontinental forces) could be used in response to an attack on the ally, the physical presence of nuclear weapons on the territory of the ally serves as strong psychological reassurance of the credibility of the nuclear threat. Further, nuclear weapons may be more likely to be used if they are based on the soil of the country in question rather than if they are located elsewhere, as they present the possibility of fighting a nuclear war that could remain limited, reducing the “suicide or surrender” credibility problem (Utgoff and Christenson 1989). The argument that nuclear weapons based locally would be more likely to be used was frequently made in Europe in the 1970s, during debates over the deployment of American intermediate range nuclear forces (INF) to Europe in response to the Soviet SS-20 missile threat (Goldstein 2000:149,149n). Forward deployed nuclear forces may also be more likely to be used if the local military authority has the physical ability to use the deployed nuclear forces, and/or if launch authority has been predelegated. Locally deployed troops may be motivated to use their nuclear forces in the face of being overrun (Bracken 1987). Last and relatedly, forward deployed weapons are more likely to be used because of “use it or lose it” pressures. In the event of war, the owner of the weapons would rather use them than lose them through capture or being overrun.

Foreign nuclear weapons deployments create two forms of entrapment risks, however. If war broke out involving the owner of the deployed weapons but not involving the weapons host state, foreign nuclear weapons based in the uninvolved host may be the target of preemptive strikes, thereby drawing the host into an unwanted war. Further, such deployments would increase the chances that a conventional conflict involving the host state would escalate to nuclear exchanges, and that the host state would not be able to prevent such escalation. In the early 1980s, the European left opposed the deployment of American INF to Europe, in part because of fears that the presence of such weapons would ensure that any conventional conflict would escalate to nuclear exchanges. On the other side of the Iron Curtain, after the Cold War ended, fears of entrapment encouraged former Warsaw Pact members in Eastern Europe to demand the removal of Soviet nuclear weapons from their territories (Kramer 1998).

In sum, a state's decision to acquire nuclear weapons will be affected by three independent variables. The first is the degree of external threat a state faces. Higher threat means a higher possibility of future conflict, in turn increasing the fear of abandonment. The second is the availability of security commitments. Such commitments, if available, can ameliorate the abandonment fears created by a highly threatening environment. The third is the state's fear of entrapment. High entrapment fears may push a threatened state to reject offered security commitments and instead build its own nuclear arsenal.

Importantly, these independent variables interact with each other, so it is important to develop nuanced predictions. First, states facing high threat and low entrapment fears both need greater security and do not exclude the option of security commitments. Hence, these states are more likely to prefer security commitments to acquiring their own nuclear arsenals. Hypothesis 1: States facing higher levels of threat and lower entrapment fears are more likely to accept offered security commitments in lieu of acquiring their own nuclear arsenals.

The above hypothesis begins from the assumption that security commitments are available. States facing high threat but without the offer of security commitments still need to increase their security and will seek nuclear weapons to do so. Entrapment fears play no role, because security commitments, the phenomena that increase entrapment fears, are not available. Hypothesis 2: States facing higher threat but without offers of security commitments are more likely to acquire nuclear weapons.

Last, entrapment fears can nullify the attractiveness of security commitments. Hence, insecure states with entrapment fears will resort to nuclear acquisition even when security commitments are offered. Hypothesis 3: States facing higher threat and having higher entrapment fears are more likely to acquire nuclear weapons, even if third party security commitments are offered.

The next two sections present quantitative and qualitative empirical tests. Unfortunately, insufficient quantitative data on entrapment fears make the quantitative testing of Hypotheses 1 and 3 infeasible. Qualitative analysis permits testing of all three hypotheses.

Quantitative Empirical Tests

This section uses quantitative methods to analyze the decisions of all states to acquire nuclear weapons from 1945 to 2000.


Dependent Variable, or “Failure” Event

For nuclear weapons acquisition, we used the data set developed by Gartzke and Kroenig (2009:154; see also Hymans 2010). Dates of acquisition of deliverable nuclear weapons are: US (1945), Soviet Union (1949), United Kingdom (1952), France (1960), China (1964), Israel (1967), South Africa (1974), India (1988), Pakistan (1990).

Independent Variable: Alliances

We used the Correlates of War (COW) data on international alliances, specifically whether or not the state in question had a defense pact with a nuclear great power (Gibler 2009). A defense pact is an agreement by which the signatories agree to intervene militarily if a signatory is attacked.

Past work has included a categorical independent variable, coded 1 if the state in question has a defense pact with a nuclear-armed state, and 0 otherwise. This approach may be flawed. It presumes that any nuclear state always extends nuclear deterrence with any alliance agreement. Most nuclear states, including Britain, France, and China, do not do so (Goldstein 2000). Other than the United States, the only nuclear power that may have actively considered extended nuclear deterrence is the Soviet Union. This article tests for the possible different extended deterrent effects of alliance with various nuclear powers.

Independent Variables: Troop Deployments

We included data on US troop deployments as a measure of third party security commitment. There are no systematic data available on the foreign troop deployments of other states. We used Department of Defense data, from 1950 forward. 1

Independent Variable: Nuclear Weapons Deployments

US nuclear weapons deployments are listed in Table 1. We did not code foreign nuclear weapons deployments of other states, due to the likely incomplete information on non-American nuclear weapons deployments. 2 We used several sources to collect information on foreign deployments of American nuclear weapons. 3 The data on the types and numbers of deployed nuclear weapons are generally incomplete, so we used a dichotomous variable coded 1 if American nuclear weapons were deployed on the territory of a country in a particular year and 0 otherwise. We coded US nuclear deployments on Greenland to be on Danish territory, as Greenland was governed by Denmark over this period. The United States deployed nuclear weapons to Okinawa after 1945 while that island was under US control. The island reverted to Japanese control in 1972, and the nuclear weapons were removed that year. We did not code visits of American nuclear-armed vessels to other countries as deployment on their territory, including the homeporting of the USS Midway at Yokosuka, Japan. We excluded deployments of nonoperational nuclear warheads lacking fissile cores. We excluded nuclear delivery vehicles deployed without nuclear warheads. There are no instances of the US temporarily deploying nuclear weapons to a foreign country during a crisis. 4

View this table:

US Foreign Deployments of Nuclear Weapons, US Defense Pacts, 1945–2000

CountryYears of DeploymentYears with Defense Pact with the United StatesWent Nuclear?
United Kingdom1954–20001949–20001952
Japan (Okinawa)19721951–2000No
South Korea1958–1991(none)No
Denmark (Greenland)1958–19651949–2000No
  • Some nuclear weapons were also deployed to US territories, including Okinawa (1954–1972), Puerto Rico (1956–1975), Guam (1951–1996), Johnston Island (1964–1971), Iwo Jima (1956–1959), and Chichi Jima (1956–1964).


US Foreign Deployments of Nuclear Weapons, US Defense Pacts, 1945–2000

Tabular Analysis

Tables 2 and 3 describe the data. In these tables, a single case is the decision of a nonnuclear state during a particular year to acquire or not acquire a nuclear weapon. During each year, a state can have nuclear weapons deployed on its territory or not, or have a defense pact with the United States or not. Across the sample of country-years, in 5.8% of the cases, a country hosted US nuclear weapons during the year in question, and in 32.3% of the cases, a country had a defense pact with the United States. In 77.7% of the country-years, a state hosted US troops, and when a state did host US troops, the average of troops stationed was 4468 (SD = 24871). Table 2 provides a striking finding: across the entire data set, no nation acquired a nuclear weapon while US nuclear weapons were stationed on its territory. Table 3 describes the data on nuclear weapons acquisitions and the presence of US defense pacts. The data are a bit less straightforward here, as all cells have nonzero quantities. Chi-squared analysis might be misleading here, because the event history structure of the data means that years in which a country possessed a nuclear weapon, other than its first year of acquisition, are not included in the sample. Hypothesis testing is best conducted with an event history model.

View this table:

US Nuclear Weapons Deployment and Nuclear Weapons Acquisitions, 1945–2000

No US Nuclear Weapons DeployedUS Nuclear Weapons Deployed
Did Not Acquire Nuclear Weapons5554342
Acquired Nuclear Weapons80

US Nuclear Weapons Deployment and Nuclear Weapons Acquisitions, 1945–2000

View this table:

Defense Pacts with the United States and Nuclear Weapons Acquisitions, 1945–2000

No Defense PactDefense Pact
Did Not Acquire Nuclear Weapons39871906
Acquired Nuclear Weapons62

Defense Pacts with the United States and Nuclear Weapons Acquisitions, 1945–2000

Multivariate Analysis

Much scholarship has used quantitative analysis to study nuclear weapons (Geller 1990; Huth 1990; Asal and Beardsley 2007; Fuhrmann 2008, 2009b; Beardsley and Asal 2009a,b; Gartzke and Jo 2009; Horowitz 2009; Rauchhaus 2009; Fuhrmann and Kreps 2010; Fuhrmann and Sechser 2011). Some quantitative research has examined the causes of nuclear proliferation (Meyer 1984), yielding mixed results on whether extended nuclear deterrence slows proliferation. Jo and Gartzke (2007) found that a nuclear ally made a state significantly less likely to acquire nuclear weapons. Phillipp Bleek (2010) found that a nuclear ally made a state less likely to explore, pursue, or acquire weapons. Kroenig (2009) found that a nuclear ally made it less likely to acquire nuclear weapons, but observed only limited statistical significance, though Kroenig (2010) found no significant relationship between nuclear alliance and nuclear proliferation. Horowitz (2010, chapter 4) and Matthew Fuhrmann (2009a) found that a nuclear ally had no effect on a state's likelihood of acquiring nuclear weapons. Singh and Way (2004) produced mixed results: with an event history model, nuclear alliance has no effect on a state's decision to explore, acquire, or pursue nuclear weapons, but with a logistic model, nuclear alliance makes a state less likely to explore or acquire nuclear weapons. Müller and Schmidt (2010) conducted some simple quantitative analysis indicating that security guarantees have no effects on proliferation decisions.

We started with Kroenig's (2009) data (which in turn built on the Singh and Way 2004 data). Kroenig includes data on several variables, including nuclear weapons possession, the existence of sensitive nuclear assistance, gross domestic product (GDP), industrial capacity, interstate rivalry, and democracy. We describe these variables below.

We use an event history model that evaluates the factors which affect the time until a state elects to acquire nuclear weapons. The “failure” event is a state's acquisition of nuclear weapons. Most past quantitative work on the causes of proliferation uses event history models, or a logistic model that amounts to an event history model (Singh and Way 2004; Fuhrmann 2009a; Kroenig 2009, 2010; Bleek 2010; Horowitz 2010). We use the Efron method of handling “ties” in duration times (Box-Steffensmeier and Jones 2004, 55).

Other Independent Variables

Sensitive Nuclear Assistance is a dichotomous variable coded 1 if the state ever received critical materials or technologies necessary for building nuclear weapons (Kroenig 2009). We also included three measures of economic development, GDP/capita, GDP/capita squared (to allow for the possibility of a nonlinear relationship), and Industrial Capacity (a dichotomous variable coded 1 if the state produces steel domestically and can generate at least 5000 megawatts of electricity). As a measure of external threat, we included the square root of the total number of Militarized Interstate Disputes (MIDs) a state has participated in over the last 10 years. 5 Some past nuclear weapons work used a dichotomous measure of rivalry as a measure of threat. We prefer MID participation, as rivalry data are built on MID activity, and a MID variable provides a more direct, nondichotomous measure of external threat. We included a measure of regime type, coded 1 if a state was democratic. 6

Multivariate Results

Table 4 presents Cox proportional hazard tests of Hypothesis 2. 7 Model 1 in Table 4 first replicates the model produced in Kroenig (2009, 171, his model 3). Here, we use Kroenig's original measure of external threat, rivalry. The results are slightly different from Kroenig's because we use the Efron method for handling duration “ties” (we were able to replicate Kroenig's results perfectly using the Breslow method for handling ties, as he did). Model 2 adds three variables, a dichotomous variable for the deployment of US nuclear weapons, a count of the number of US troops deployed, and a dichotomous variable for the presence of a defense pact with the United States. 8 Model 2 uses our MID count threat variable. 9 Model 2 removes Kroenig's dichotomous alliance variable.

View this table:

Cox Event History Analysis of Causes of Nuclear Weapons Acquisition, 1945–2000

Failure EventModel 1Model 2Model 3
Sensitive Nuclear Assistance2.12 ** (0.862)1.25 (1.31)1.15 (1.33)
Civil Nuclear Agreements0.0442 (0.0430)
GDP/cap0.000621 ** (0.000237)0.000546 * (0.000310)0.000704 (0.00051)
GDP/cap squared−0.0000000568 ** (0.0000000212)−0.0000000381 ** (0.0000000155)−0.0000000489 (0.0000000312)
Industrialization3.58 *** (0.523)3.23 *** (0.595)3.25 *** (0.607)
Rivalry2.38 * (1.26)
MIDs (square root)0.720 * (0.328)0.770 * (0.364)
Democracy0.112 * (0.0553)0.125 (0.0861)0.111 (0.0928)
Defense Pact with Any Nuclear Power−1.72 * (0.954)
Defense Pact with the United States−1.73 (1.19)−1.94 * (1.06)
US Nuclear Deployment−38.2 *** (1.01)−43.7 (no SE estimate)
US Troops−0.00000110 (0.000000781)−0.000000480 (0.00000715)
Log pseudo-likelihood−19.06568−17.421038−17.016907
  • Coefficients rather than hazard rates are reported.

  • All significance tests are one-tailed.

  • ***p < 0.001; **p < 0.01; *p < 0.05. Efron method for handling ties was used.


Cox Event History Analysis of Causes of Nuclear Weapons Acquisition, 1945–2000

In Model 2, neither the US Defense Pact nor Troops variables are statistically significant. However, the Nuclear Deployments variable is highly statistically and substantively significant. This is unsurprising, as there are no cases of a state going nuclear when there were US nuclear weapons deployed on its soil. The Nuclear Assistance variable is not statistically significant. The MID count variable is significant. If we instead use a variable that is limited to MIDs with nuclear-armed countries, the results remain mostly unchanged. 10 In Model 3, we add the Fuhrmann (2009a, 2009b) measure of Civil Nuclear Agreements, and it is insignificant. 11

The results support Hypothesis 2. Higher threat coupled with an absence of at least some security guarantees increase abandonment fears and make nuclear proliferation more likely. Note that though the nonexistence of episodes of a country hosting US nuclear weapons while acquiring its own nuclear weapons suggests that US nuclear deployments may be sufficient to stop nuclear acquisition, such deployments are not necessary to prevent a country from acquiring nuclear weapons. Countries such as Morocco host US nuclear weapons even though they face low threat environments and are hence unlikely to go nuclear.

As Table 2 indicates, there are no instances of a state going nuclear while US nuclear weapons were deployed on its soil. This is a data condition known as separation, occurring when one of the cells in a 2 × 2 table of an independent variable and the dependent variable is empty (to some degree, the independent variable “perfectly” predicts the dependent variable). Separation makes estimating marginal effects quite difficult. Different statistical packages treat separation differently and can arrive at widely different coefficient estimates (Zorn 2005). Relatedly, though Model 2 in Table 3 presents a full set of results, testing slightly different specifications often produces coefficient estimates but no standard error estimates for the nuclear deployments variable. The lack of standard error estimates in some models suggests a separation problem and is not evidence that the null hypothesis should not be rejected. 12

The separation issue notwithstanding, the cross-tabulation in Table 2 combined with the multivariate results in Table 4 indicates that deployment of American nuclear weapons reduces the likelihood that a state will acquire nuclear weapons. But what of pursuit of nuclear weapons, prior to acquisition? We examined pursuit as a dependent variable by recoding the failure event, changing it from the year of acquisition to the first year that a state decided to pursue nuclear weapons. We used the revised Singh and Way (2004) data set on nuclear weapons pursuit. 13 They (866) define a state as pursuing a nuclear weapon if it has taken any concrete steps toward acquiring nuclear weapons, including “political decision by cabinet-level officials, movement toward weaponization, or development of single-use, dedicated technology.”

The results of this analysis are displayed in Model 4 of Table 5. The Troops and Defense Pact variables remain statistically insignificant, and the US Nuclear Deployments variable is also insignificant. If we instead use the Bleek (2010, 169) nuclear pursuit data, the results are essentially similar, with the exception that the Nuclear Assistance variable is no longer statistically significant. 14 The Deployments variable is probably insignificant because two states are coded as beginning their pursuit of nuclear weapons while US nuclear weapons were deployed on their soil, South Korea in 1970 and Taiwan in 1967. In Model 5, we add the Civil Nuclear Agreements variable. It is insignificant, and the results do not change.

View this table:

Additional Cox Event History Analysis of Causes of Nuclear Weapons Pursuit and Acquisition, 1945–2000

Model 4Model 5Model 6Model 7
Failure EventPursuitPursuitAcquisitionAcquisition
Sensitive Nuclear Assistance1.65 ** (0.679)1.57 ** (0.654)1.23 (1.40)1.15 (1.43)
Civil Nuclear Agreements0.0470 (0.0665)0.0441 (0.0414)
GDP/cap0.000393 ** (0.000166)0.000415 ** (0.000147)0.000542 (0.000349)0.000697 (0.000533)
GDP/cap squared−0.0000000345 * (0.0000000157)−0.0000000401 * (0.0000000183)−0.0000000380 * (−0.0000000171)−0.0000000487 (0.0000000316)
Industrialization2.26 *** (0.388)2.04 *** (0.479)3.21 *** (0.611)3.23 *** (0.626)
MIDs (square root)1.16 *** (0.222)1.15 *** (0.222)0.717* (0.342)0.765* (0.379)
Democracy−0.0130 (0.0327)−0.0131 (0.0306)0.125 (0.0947)0.111 (1.02)
Defense Pact with the United States−0.110 (0.768)−0.206 (0.733)−1.75 (1.21)−1.98 * (1.01)
Defense Pact with USSR0.107 (1.27)0.163 (1.23)
Defense Pact with China−41.1 (no SE estimate)−40.6 (no SE estimate)
US Nuclear Deployment0.486 (1.02)0.378 (1.07)−43.2 (no SE estimate)−44.1 (no SE estimate)
US Troops−0.0000168 (0.0000155)−0.0000142 (0.0000146)−0.000000801 (0.00000737)−0.0000000707 (0.00000694)
Log pseudo-likelihood−50.837661−50.396384−17.385415−16.987527
  • Coefficients rather than hazard rates are reported.

  • All significance tests are one-tailed.

  • ***p < 0.001; **p < 0.01; *p < 0.05. Efron method for handling ties was used.


Additional Cox Event History Analysis of Causes of Nuclear Weapons Pursuit and Acquisition, 1945–2000

South Korea and Taiwan may be the exceptions that prove the rule. They appear to have pursued nuclear weapons even in the context of a third party security commitment because they perceived the credibility of that commitment to be declining (for more on South Korea, see next section). Beyond this specific point, the results perhaps highlight the diplomatic and geopolitical costs of nuclear acquisition, relative to nuclear pursuit. The results are consistent with the proposition that states pay higher costs when they acquire nuclear weapons in comparison with pursuing nuclear weapons. Nuclear weapons pursuit can occur in secret, virtually always in the realm of plausible deniability. Nuclear weapons acquisition, however, is usually a public event, often a nuclear test. Common knowledge that a state has gone nuclear can mean higher costs, in terms of diplomatic isolation, being targeted for economic sanctions and international opprobrium, and stimulating the nuclear programs of rivals and neighbors. States may be less willing to pay those costs if foreign nuclear weapons deployments offer the security benefits that an indigenous nuclear arsenal would provide. However, because the diplomatic and geopolitical costs of nuclear weapons pursuit are lower, states may be more open to nuclear pursuit short of acquisition, even if American nuclear forces have been deployed. This may have been the case for South Korea and Taiwan, two states that sought to pursue nuclear arsenals secretly as hedges in the face of the decline in the credibility of the American security guarantee.

The quantitative analysis does not support the propositions that a defense pact with a nuclear ally significantly decreases the likelihood that the state will go nuclear, or that basing troops in a country will significantly decrease its likelihood of going nuclear. The insignificant relationship of nuclear weapons acquisition with defense pacts deserves a bit more discussion, as it runs against the findings in some earlier studies that alliance with a nuclear-armed state reduces the likelihood that a state will go nuclear. To explore this further, consider Model 6 in Table 5. This model includes three different alliance variables: defense pact with the United States, defense pact with a nuclear USSR, and defense pact with a nuclear China. 15 Neither the United States or Soviet defense pact variables are statistically significant, but the China defense pact variable is significant and negative. The statistical analysis fails to produce standard error estimates, because that variable suffers from separation, as there are no cases over this period of a state with a defense pact with China acquiring nuclear weapons. When we reran the analysis with data from the Alliance Treaties Obligations and Provisions data (ATOP), the results were similar, with the United States and Soviet alliance variables being statistically insignificant, and the China alliance variable being significant and positively signed (Leeds, Ritter, McLaughlin Mitchell and Long 2002). Model 7 of Table 5 includes the Civil Nuclear Agreements variable, and it remains insignificant.

This finding apparently showing that a defense pact with China makes nuclear proliferation less likely is curious, as we should expect that United States and Soviet defense pacts should have stronger nonproliferation effects than Chinese agreements, given the greater size and capabilities of US and Soviet nuclear forces, and the greater global engagement of American and Soviet foreign policies. The likely explanation for the significant effect of a Chinese defense pact is spurious correlation for one case. China had one defense pact with one non-nuclear state after 1964 (when China went nuclear), North Korea. North Korea did not acquire nuclear weapons by 2000, the end of the data set. However, closer examination casts doubt on the inference that the Chinese defense pact dissuaded North Korea from going nuclear. North Korea pursued nuclear weapons as early as the 1960s (Oberdorfer 1997:252–53). North Korean pursuit of nuclear weapons is perhaps not surprising, as it appears that China neither had the intention or capability of using its nuclear weapons to extend deterrence. China has maintained a declaratory stance of nuclear no first use, meaning it would not use nuclear weapons in response to a conventional attack on North Korea. China has never alerted its nuclear arsenal as a means of signaling a deterrent threat (Lewis 2007). It has maintained a small nuclear arsenal, in the low tens of warheads, to maintain “the minimum means of reprisal if attacked by the imperialist with nuclear weapons” (quoted in Lewis 2007:1). Indeed, in one of its very few public statements about its nuclear strategy, China in 1964 explicitly excluded the possibility of using nuclear weapons in defense of North Korea: “China is developing nuclear weapons for defense and for protecting the Chinese people from US threats to launch a nuclear war” (italics added; quoted in Lewis 2007:60). Lastly, if the analysis was extended forward to 2006, when North Korea detonated a low yield nuclear device while still a Chinese ally, the coefficient might become statistically insignificant. In short, the observed correlation between alliance with nuclear powers and proliferation turns on a single case, and close observation reveals that the correlation in that case is likely spurious.

Endogeneity and Causality

A possible critique of the nuclear weapons deployment finding is that the observed correlation between nuclear weapons deployments and nuclear non-proliferation may not justify the inference that nuclear deployments prevent proliferation. If the United States deploys nuclear weapons to countries especially unlikely to go nuclear, then the observed negative relationship between deployments and nuclear acquisition might be spurious. However, because a primary American goal for deploying American nuclear forces is to reduce proliferation motives, the expectation is that the United States would be more likely to deploy nuclear forces to those countries most likely to acquire nuclear weapons. That is, we would expect that because the United States sends its nuclear weapons to countries most likely to go nuclear. This would if anything bias the analysis toward finding no relationship between US nuclear deployments and nuclear acquisition. Hence, we can be especially confident that the finding that nuclear deployments are negatively associated with nuclear acquisition is not spurious, and not contaminated by endogeneity bias. 16

Individual cases

This section describes categories of cases representing various outcomes predicted by the theory and hypotheses.

Abandonment Dominates Entrapment: Security Commitments Replace Nuclear Programs

There are several states that had greater abandonment fears than entrapment fears and accepted third party security commitments in lieu of pursuing nuclear weapons (Hypothesis 2). For example, Norway's mild interest in developing its own nuclear arsenal was likely dissolved by its entry into NATO (Forland 1997, 13).

Perhaps more prominently, West Germany faced a severe external threat from the Soviet Union in the 1950s and possessed a scientific and technological base sufficient to construct a nuclear weapon. It also feared abandonment, that in the event of another war in Europe, the United States might hesitate before committing to its defense and risk the destruction of the American homeland. German entrapment fears were lower because any new world war would very likely begin with a Soviet invasion of West Germany (Trachtenberg 1991:119).

The United States sought to reduce West German fears of abandonment sufficiently to dissuade West Germany from acquiring independent national nuclear forces. It arranged West Germany's 1954 commitment not to produce nuclear weapons as part of a broader agreement that included West German entry into NATO, deployment of American troops to German soil, and American commitment to German defense, the last including the deployment of American nuclear weapons to Germany as early as 1955 (Kelleher 1975:9–10). In the late 1950s, several possible German nuclear arrangements were discussed, but by 1963, West German nonnuclear status was reaffirmed as part of a broader arrangement that included Soviet acceptance of the status of West Berlin and enduring American military deployments on German soil (Trachtenberg 1999:397–98).

South Korea also fits into this category of security commitments discouraging proliferation. 17 The 1953 Mutual Defense Treaty between the United States and South Korea provided the legal basis for a long-term US security commitment to South Korea. 18 American forces deployed to South Korea affirmed America's commitment to South Korea's defense. American nuclear forces were first deployed to South Korea in 1958. South Korea accepted its status as a nonnuclear state, signing the Nuclear Non-Proliferation Treaty in 1968. South Korea's entrapment fears were lower, as its agreement with the United States did not require South Korean assistance following an attack on the United States, the agreement was bilateral rather than multilateral, and the American nuclear forces deployed to its territory were shorter range, lowering Soviet preemption incentives.

However, doubts in America's security commitments encouraged South Korea to reconsider its nonnuclear status. In 1969, President Richard Nixon announced the “Guam Doctrine,” calling on America's Asian allies to take on more responsibility for their own defense. The following year, the United States began to replace US forces in South Korea with more military aid. Soon thereafter, the United States opened diplomatic relations with China and withdrew from South Vietnam.

Consistent with Hypothesis 2, South Korean interest in a nuclear program grew alongside a perceived weakening of American commitments and a growing fear of abandonment. The South Korean government created the Agency for Defense Development in 1970, an organ that soon recommended acquiring nuclear weapons. The leadership agreed and began secretly to lay the groundwork for a nuclear weapons program. In June 1975, the South Korean President declared publicly that South Korea might acquire nuclear weapons if the American nuclear guarantee were removed.

In late 1975, the United States was able to compel South Korea to back down from its nuclear ambitions, threatening to terminate American civilian nuclear assistance as well as perhaps the fundamentals of the American-South Korean relationship if South Korea's nuclear program proceeded. However, President Carter's 1977 declared intention to withdraw American forces from South Korea caused the South Korean government to declare that it would procure nuclear weapons if American forces were withdrawn. Carter's retreat on his pledge dissuaded South Korea from advancing its nuclear weapons program. In 1991, American nuclear forces were withdrawn from South Korea, but this occurred in the context of a joint US-South Korean move to denuclearize the Korean peninsula and increased South Korean interest in accommodation with the North.

Abandonment Fears Plus Insufficient Security Commitments Spur Nuclear Programs

The theory predicts that threatened states with high abandonment concerns and low entrapment fears will desire third party security guarantees (Hypothesis 1). If such guarantees are unavailable, then threatened states may seek their own nuclear weapons programs as means of redressing the threat and solving the abandonment problem (Hypothesis 2). Several states fit this description. Israel appears to have initiated its nuclear program in the late 1950s in the context of a highly threatening environment coupled with nonexistent third party security commitments. Israel had fought two major wars with its Arab neighbors in less than a decade (1948 and 1956), the latter ending with American intervention to halt the Anglo-French-Israeli advance. For Prime Minister David Ben-Gurion, the leader who decided to initiate Israel's nuclear weapons program, “the lesson from the Holocaust was that a small Jewish state without a formal alliance with an outside world power would have to create its own existential insurance policy” (Cohen 2010:36).

The theory also sheds light on South Africa's decision to acquire nuclear weapons. Concern about declining security commitments from the West coupled with growing concerns about security threats pushed South Africa to build a nuclear weapon. A secret 1975 memo written by the South African Chief of the Defense Staff “argued that a significant nuclear threat to South Africa had emerged, justifying the acquisition of a nuclear weapons capability. The threat envisaged was that a hostile African nation might acquire a nuclear weapon from China, and that a United States pursuing East-West détente could not be counted upon to come to South Africa's aid” (Harris, Hatang and Liberman 2004:461).

Pakistan is another example of a state with high abandonment fears and low entrapment fears seeking security commitments, but pursuing its own nuclear weapons when such commitments were unavailable. By the middle 1970s, Pakistan faced a highly threatening security environment. It had fought two major wars with India within less than a decade, in 1965 and 1971. The threat escalated with India's 1974 “peaceful nuclear explosion.” Pakistan and the United States had a bilateral defense pact dating back to 1959. However, the Pakistanis had doubts about the depth of American commitments. The United States had sanctioned Pakistan in 1965 following the Pakistani use of American weapons in the war against India. The United States had refused to aid Pakistan in 1971 during the Bangladesh War. With that in mind, in the wake of the Indian nuclear test, Pakistan dispatched its ambassador to visit Secretary of State Henry Kissinger, formally requesting that the United States put Pakistan under its nuclear umbrella. Kissinger refused, pushing Pakistani prime minister Zulfikar Ali Bhutto to initiate a Pakistani nuclear weapons program, to “save the nation” (Levy and Scott-Clark 2007:12–14).

Abandonment Fears and Entrapment Fears Encourage Nuclear Programs over Foreign Nuclear Deployments

States facing abandonment and entrapment fears may address perceived insecurity by acquiring their own nuclear weapons rather than rely on external security commitments (Hypothesis 3). 1950s France is a good example. Fearing both abandonment and entrapment, France opposed American nuclear weapons deployments and pursued its own nuclear arsenal. The French move to nuclear weapons was pushed forward by abandonment fears exacerbated by the 1950s Dienbienphu, Algeria, and Suez War episodes, all of which indicated to Paris that the United States would not steadfastly defend French global interests (Goldstein 2000:188–91; for a critique, see Hymans 2006:94). France also became increasingly concerned in the 1950s that the growth in the Soviet arsenal undermined the credibility of the American extended nuclear deterrent (Harrison 1981:36). When Charles de Gaulle became premier in June 1958, one of his first foreign policy decisions was to declare that he would not allow the placement of any non-French nuclear weapons on French soil. This was coupled with a decision to accelerate the French nuclear program, aiming to conduct France's first nuclear test in 1960. De Gaulle conveyed the decision not to host foreign nuclear weapons to Secretary of State John Foster Dulles when they met in July 1958. Dulles hoped to dissuade the French from developing an independent nuclear force and offered to deploy American nuclear forces to French soil as a means of strengthening the American extended deterrent, hopefully reducing France's perceived need for its own nuclear arsenal. De Gaulle rebuffed Dulles' suggestion, in part because of entrapment fears (Challener 1994:152–54; Reyn 2010:35, 54).

Late 1940s Britain also confronted abandonment and entrapment concerns. Britain's January 1947 decision to build its own nuclear arsenal was driven in part by abandonment concerns, specifically the growing Soviet threat coupled with the lack of a formal American security commitment (at the time), the small American nuclear arsenal, and the lack of a strategic American bomber capability. In the late 1940s, Britain disagreed with the American suggestion that all nuclear weapons and production facilities be located in North America, on the basis of the abandonment concern that in war the use of such weapons would not be in British control, and that such weapons might not get deployed to Britain in time (Gowing 1974: volume 1:286). The security concerns were related to the belief by some that Britain needed an independent nuclear force to remain a great power (Peterson 1997:7). Over the next 5 years, the expansion of the American nuclear arsenal and formation of NATO were insufficient to dissuade Britain from continuing on this path. The outbreak of the Korean War highlighted possible strategic disagreements between the United States and Britain, and the need for an independent British nuclear deterrent. Britain wanted its own nuclear arsenal to permit nuclear use in the event of American abandonment (Gowing 1974: volume 1:309). Britain conducted its first nuclear test in 1952.

Entrapment fears shaped Anglo-American discussions about the deployment of US nuclear weapons to Britain during this time period. Some in Britain feared that the United States might decide to use its nuclear weapons ahead of British approval, prematurely dragging Britain into a nuclear war (Baylis 1995: chapters 3–4). These entrapment fears grew in the Korean War, when Britain pressured the United States not to use nuclear weapons against North Korea or China, fearing escalation (Risse-Kappen 1995: chapter 3). Some also feared that American air bases in Britain might become a target for Soviet attacks in an emerging US-Soviet conflict. Britain was able to reduce its entrapment fears by securing American commitments as early as 1951 that the American use of air bases in Britain, from which nuclear weapons would be launched, would be a joint Anglo-American decision (Duke 1987:77). With this commitment in hand, Britain eventually accepted the deployment of American nuclear weapons to British soil, in 1955.


The cases illustrate the patterns predicted in Hypotheses 1–3 and more broadly point to the importance of understanding how the interplay of abandonment fears, entrapment fears, and the supply of third party security commitments affects proliferation decisions. The cases also suggest other inferences. First, consistent with the null quantitative alliance findings, the mere presence of security commitments may not be sufficient to reassure states with abandonment fears. NATO was insufficient to allay French fears of abandonment because of damaged US credibility. The US-South Korea treaty was insufficient to calm South Korean fears in the 1970s. These patterns are consistent with the view that international agreements are limited in their ability to bind the actions of states.

Second, nuclear weapons deployments reassure more effectively than alliances, perhaps because foreign nuclear deployments are more credible than written commitments. Nuclear weapons deployed to an invasion target are relatively likely to be used, whereas a target's ally might renege on a written commitment. As noted in the introduction, the Baltic States and South Korea desire the deployment of American nuclear forces on their soil to address perceived threats, despite the fact that all four states are American allies. That said, troop deployments do not have the reassuring effects of nuclear weapons. There is little uncertainty about the escalation effects of nuclear weapons, but an attacker might believe it could defeat an ally's deployed troops in battle.

Third, alliances may be a more complex nonproliferation tool than allowed by the theory presented here. The South Korea case indicates that third parties can use alliances to persuade their partners to forego proliferation. Such tools were used in the 1970s to prevent Taiwan from going nuclear, as well (Hersman and Peters 2006).

Fourth, the cases indicate when entrapment fears might be greater or lesser. Smaller states in bilateral alliances, like South Korea, Pakistan, and Taiwan, are less likely to fear entrapment. These states are unlikely to be dragged into conflict they would otherwise be disinterested in. States in multilateral alliances likely to be attacked first in a larger war, such as West Germany, also have lower entrapment fears. Conversely, states in multilateral alliances that are not on the front lines of conflict, such as France and Britain, are more likely to fear entrapment. They face greater risk of being drawn into a war they might otherwise avoid. Bilateral vs. multilateral aside, the deployment of longer range nuclear forces may increase fears of entrapment via preventive or preemptive nuclear attack, as was the case in Britain. Shorter range nuclear forces, such as were deployed to South Korea and Taiwan, may be less tempting targets for preventive or preemptive attack, reducing entrapment concerns.


Abandonment fears, entrapment fears, and the availability of third party security commitments are all important security factors that affect a state's decision to acquire nuclear weapons. The quantitative results for the article present a new finding: no state hosting foreign deployed nuclear weapons has ever acquired its own nuclear weapons. Nuclear weapons deployments do not appear to slow the pursuit of nuclear weapons short of acquisition, however, as threatened states hosting foreign deployed nuclear weapons may pursue nuclear weapons as insurance in case foreign deployed nuclear weapons are withdrawn. Higher threat makes a state more likely to acquire nuclear weapons.

The quantitative results generally show that alliances with nuclear powers do not make a state less likely to go nuclear. Indications of a positive relationship between alliance and nonproliferation may be driven by a spurious relationship between the China-North Korean alliance and North Korean nonproliferation. Foreign troop deployments also failed to make a state less likely to go nuclear. Outside of security factors, perhaps the most salient empirical finding is that nuclear assistance does not make it more likely that a state will go nuclear.

The qualitative analysis provided more support for the theory, describing states that forewent nuclear weapons because of security commitments (South Korea, West Germany, and Norway), that acquired nuclear weapons at least in part because of the absence of sufficient security commitments (South Africa, Israel, and Pakistan), and that went nuclear in part because of concerns about the entrapment dangers of foreign nuclear weapons deployments (France and Britain). They also provided nuance to the null alliance results in the quantitative section, demonstrating that sometimes alliances can reassure a state enough to push it away from nuclear acquisition. However, formal alliances will fail to reassure if a state has doubts in its ally's credibility. Further, states come to doubt an ally's credibility not because of a record of noncompliance with the formal terms of the alliance, but rather because of other, less direct signals about an ally's reliability. This suggests the general conclusion that the effectiveness of formal alliances is perhaps more fragile than previously assumed.

These findings have important policy implications. If a partner state is willing, American nuclear deployments could be an effective mechanism to reassure countries like South Korea and the Baltic States, reducing those states' incentives for acquiring nuclear arsenals. Enhanced US security commitments to Saudi Arabia could inoculate that country from responding in kind to Iran's nuclear program. The findings indicate that effective reassurance may require US nuclear deployments, as forming alliance agreements or deploying US troops might not be enough to steer a state away from going nuclear. The null finding on nuclear assistance also indicates that controlling nuclear trade through instruments such as the Nuclear Suppliers Group might be of limited effectiveness.

Caution is in order, however. Such deployments might fuel arms races, make nuclear accidents and terrorism more likely, and inflame anti-Americanism. Perhaps more importantly, foreign nuclear deployment is likely not a feasible option for addressing rogue, anti-American nuclear aspirants like Iran. To deal with potential nuclear proliferators like Iran, the United States and the rest of the international community must consider other options, such as diplomacy and economic sanctions.


  • * Earlier versions of this article were presented at the September 2011 Concepts and Analysis in Nuclear Strategy conference in Washington, DC, and the December 2011 UC San Diego “Politics of Nuclear Weapons in the 21st Century” in Washington, DC. For research assistance, thanks to Andrew Ratto. For comments, thanks to Kyle Beardsley, Matthew Fuhrmann, Michael Horowitz, Drew Linzer, and Todd Sechser.

  • 1 This data, provided by the Department of Defense, was compiled by Tim Kane, and is available at http://www.heritage.org/Research/Reports/2006/05/Global-US-Troop-Deployment-1950–2005 <downloaded May 12, 2011>. Data are not available for the foreign troop deployments of other nations.

  • 2 There is some evidence that Britain deployed nuclear weapons at one point to Cyprus and the Soviet Union deployed nuclear weapons at various points to some Eastern European countries and Cuba (Fuhrmann and Sechser 2011). Since none of those host countries acquired nuclear weapons, including data on the foreign deployment of non-US nuclear weapons would strengthen the finding that foreign deployment of nuclear weapons is negatively correlated with nuclear proliferation.

  • 3 Sources available from author.

  • 4 Our data set on foreign nuclear deployments parallels that of another new data set on nuclear weapons deployment (Fuhrmann and Sechser 2011).

  • 5 We used MID start year to code year of participation. We used the square root to reduce the effect of outliers.

  • 6 Kroenig in some models included measures of Economic Openness and Liberalization. He did not find them to be statistically significant, so we did not include them here. We do not include a measure of NPT membership, as such a measure is likely endogenous to a decision to acquire nuclear weapons.

  • 7 The partial correlations among the independent variables are not worrisomely high. For example, the correlation between Defense Pact and Troops is .35, the correlation between Nuclear Deployments and Defense Pacts in .28, and the correlation between Troops and Nuclear Deployments is .13.

  • 8 Analysis revealed that the proportional hazard assumption was not violated (Box-Steffensmeier, Reiter and Zorn 2003).

  • 9 If we run Model 2 with rivalry instead of the MID count variable, the rivalry variable is insignificant, and the other results are largely unchanged.

  • 10 The exceptions are that the nuclear assistance variable becomes statistically significant, and no standard error estimates are produced for US nuclear deployments, likely because of separation.

  • 11 It remains insignificant if we use the Fuhrmann (2009a,b) measure of all nuclear agreements or all nonsafety nuclear agreements, either as count variables or dichotomous dummy variables.

  • 12 Robustness tests altering specifics of the research design upheld the core results, that nuclear deployments make proliferation less likely, that defense pacts do not make a state significantly less likely to go nuclear, and that troop deployments do not make nuclear acquisition less likely. Robustness tests included: using a Weibull model; dropping all the control variables; replacing the US defense pacts variable with Kroenig's variable of alliance with any nuclear power; using different measures of US troop deployments; coding India as going nuclear in 1974; using ICB data to code the rivalry variable instead of MID data; recoding the alliance variable to include ententes as well as defense pacts; recoding the US-South Korea agreement as a defense pact rather than an entente.

  • 13 The original codings were revised in early 2011. http://falcon.arts.cornell.edu/crw12/documents/Nuclear%20Proliferation%20Dates.pdf <downloaded October 4, 2011>.

  • 14 Another approach would be to look at the causes of nuclear acquisition just among states engaged in nuclear pursuit. The limits of such an analysis is that it suffers potential selection bias, as it ignores the process in which a state chooses to pursue nuclear weapons. It also substantially reduces the number of subjects, to 18. That said, in analysis of the causes of acquisition among state that pursue, foreign nuclear deployments remains significant, but many other results change: the troops variable becomes significant but signed in the wrong direction, the MID variable becomes insignificant, the US alliance variable becomes significant and signed in the predicted direction, and the GDP variables become insignificant (results not shown).

  • 15 India, Israel, Pakistan, Britain, France, and South Africa did not extend deterrence over other states. British and French defense pacts are generally subsets of American alliance networks.

  • 16 Perhaps threat is positively correlated with the likelihood of acquiring nuclear weapons, the likelihood of troop deployments, and the likelihood of forming an alliance, and though there is a real, negative causal relationship between alliances and troops and acquiring nuclear weapons, that effect is washed out in the analysis. However, other studies have demonstrated that in the post-1945 period common threat does not make two states more likely to ally with each other (Lai and Reiter 2000), and a state is not significantly more likely to host US troops if it faces higher levels of threat (Biglaiser and DeRouen 2009).

  • 17 The following paragraphs draw on Solingen 2007, 83; Reiss 1988; chapter 3; Pollack and Reiss 2004; Hersman and Peters 2006; http://www.gwu.edu/∼nsarchiv/NSAEBB/NSAEBB87/ <downloaded February 7, 2012>.

  • 18 It is coded as a consultation agreement by COW, and as a defense pact by ATOP.


View Abstract