Assessing State Nuclear Weapons Proliferation: Using Bayesian Network Analysis of Social Factors

15
DRAFT Assessing State Nuclear Weapons Proliferation: Using Bayesian Network Analysis of Social Factors 1 By: Garill Coles, Alan Brothers, Jarrod Olson 2 , and Paul Whitney ABSTRACT: A Bayesian network (BN) model of social factors can support proliferation assessments by estimating the likelihood that a state will pursue a nuclear weapon. Social factors, including political, economic, nuclear capability, security, and national identity and psychological factors may play as important a role as more physical factors in whether a State pursues nuclear weapons. This paper illustrates how Bayesian reasoning can be used to combine evidence that supports proliferation assessment for a generic case of a would-be proliferator State . Theories and analysis by political scientists can be leveraged in a quantitative and transparent way to indicate proliferation risk. BN models facilitate diagnosis and inference in a probabilistic environment using a network of nodes and acyclic directed arcs between the nodes whose connections, or absence of, indicate probabilistic relevance, or independence. We propose a BN model that uses information from both traditional safeguards and the strengthened safeguards associated with the Additional Protocol to indicate countries with a high risk of proliferating nuclear weapons. Such a model could be used in a variety of applications such a prioritization tool and as a component of state safeguards evaluations. This paper will discuss the benefits of BN reasoning, the development of Pacific Northwest National Laboratory’s (PNNL) BN state proliferation model, and how the model could be employed as an analytical tool. Introduction In 1962, John F. Kennedy famously predicted that the world faced a cascade of nuclear weapons proliferation as nuclear technology and knowledge continued to spread. This prediction mirrored the common belief in the “Technological Imperative” common at the time, where acquisition of nuclear energy technology would inevitably yield nuclear weapons technology. Fortunately, this proved to be an inaccurate forecast. Political scientists have since continued to examine the root causes of proliferation and build frameworks for understanding nuclear weapons proliferation. However, one commonly acknowledged shortcoming of existing studies of proliferation is that they rely on historical precedent. They identify when a state is following the same path as historic proliferators, but they may not be able to predict the next wave of proliferators. In 2009, Alexander Montgomery and Scott Sagan identified key shortcomings of proliferation studies. 3 They noted that data are scarce about nuclear weapons and it is difficult to 1 PNNL-SA-72126 2 Contact author: 1100 Dexter Ave N. Suite 400, Seattle, WA 98109. [email protected] 3 Montgomery, Alexander H., Scott Sagan. The Perils of Predicting Proliferation. Journal of Conflict Resolution, 4 February, 2009, http://online.sagepub.com/ doi: 10.1177/0022002708330581

Transcript of Assessing State Nuclear Weapons Proliferation: Using Bayesian Network Analysis of Social Factors

DRAFT

Assessing State Nuclear Weapons Proliferation: Using Bayesian

Network Analysis of Social Factors1 By: Garill Coles, Alan Brothers, Jarrod Olson2, and Paul Whitney

ABSTRACT:

A Bayesian network (BN) model of social factors can support proliferation assessments by estimating the

likelihood that a state will pursue a nuclear weapon. Social factors, including political, economic,

nuclear capability, security, and national identity and psychological factors may play as important a role

as more physical factors in whether a State pursues nuclear weapons. This paper illustrates how

Bayesian reasoning can be used to combine evidence that supports proliferation assessment for a

generic case of a would-be proliferator State . Theories and analysis by political scientists can be

leveraged in a quantitative and transparent way to indicate proliferation risk. BN models facilitate

diagnosis and inference in a probabilistic environment using a network of nodes and acyclic directed

arcs between the nodes whose connections, or absence of, indicate probabilistic relevance, or

independence. We propose a BN model that uses information from both traditional safeguards and the

strengthened safeguards associated with the Additional Protocol to indicate countries with a high risk of

proliferating nuclear weapons. Such a model could be used in a variety of applications such a

prioritization tool and as a component of state safeguards evaluations. This paper will discuss the

benefits of BN reasoning, the development of Pacific Northwest National Laboratory’s (PNNL) BN state

proliferation model, and how the model could be employed as an analytical tool.

Introduction

In 1962, John F. Kennedy famously predicted that the world faced a cascade of nuclear weapons

proliferation as nuclear technology and knowledge continued to spread. This prediction mirrored the

common belief in the “Technological Imperative” common at the time, where acquisition of nuclear

energy technology would inevitably yield nuclear weapons technology. Fortunately, this proved to be an

inaccurate forecast.

Political scientists have since continued to examine the root causes of proliferation and build

frameworks for understanding nuclear weapons proliferation. However, one commonly acknowledged

shortcoming of existing studies of proliferation is that they rely on historical precedent. They identify

when a state is following the same path as historic proliferators, but they may not be able to predict the

next wave of proliferators. In 2009, Alexander Montgomery and Scott Sagan identified key shortcomings

of proliferation studies.3 They noted that data are scarce about nuclear weapons and it is difficult to

1 PNNL-SA-72126

2 Contact author: 1100 Dexter Ave N. Suite 400, Seattle, WA 98109. [email protected]

3 Montgomery, Alexander H., Scott Sagan. The Perils of Predicting Proliferation. Journal of Conflict Resolution, 4

February, 2009, http://online.sagepub.com/ doi: 10.1177/0022002708330581

DRAFT

code independent variables like prestige, bureaucratic power and normative effects of the international

system. In addition, they maintain that most studies tend to provide insights that most experts already

know and understand, thus failing to provide additional information to policy makers.4

Understanding this landscape, PNNL undertook a research and development project to assess the utility

of social modeling for proliferation assessment. We chose Bayesian Nets (BNs) as our modeling

environment because they have the ability to integrate statistical data with expert judgment,5 where

data are unavailable, in order to build a stronger predictive model. BN models can incorporate the

theories and analyses of political scientists and academics. The qualitative structure of a BN is intuitive

and easily understood, and they can be quantified to provide a useful metric for the detection and

deterrence of nuclear weapons proliferation.

This paper will discuss PNNL’s development of a predictive BN model and some of the challenges the

team has encountered. The first section covers background material on nuclear proliferation analysis;

the second section covers information about BN models; the third section discusses the PNNL model

development and the final section covers the next steps that the project team plans to take.

Background

In 1970, the proliferation paradigm shifted with the development of a legal and normative treaty known

as the Nuclear Non Proliferation Treaty (NPT). The NPT introduced a system of constraints and

normalized an inspection regime for nuclear energy programs, building transparency into the secretive

world of nuclear weapons development and enabling the peaceful spread of nuclear energy without the

negative side-effects of nuclear weapons proliferation. Nonetheless, the system has not worked

perfectly and it has been challenged on a regular basis by states with negative intentions.

For decades authors have attempted to answer the question, “who will be next?” with a series of

different approaches. In the early years of nuclear proliferation literature, authors like Epstein pondered

the question with academic reasoning based on well-known theories of political science.6 Over time, this

type of work developed into qualitative case-study research and eventually blossomed into quantitative

research utilizing statistics to test theories of proliferation.

Around the advent of nuclear weapons, there was a common belief that nuclear energy and nuclear

weapons were undeniably linked. Assumptions like this led to President Kennedy’s famous prediction in

1962. As new nuclear weapons development ebbed following the Chinese test despite the continuing

spread of nuclear energy technology, theorists were challenged to explain the phenomenon. Working

4 Ibid.

5 “Expert judgment” for the purpose of this paper refers to a structured, methodological approach to gathering

data from experts and incorporating it into the models. In some cases, these data are preliminary and will be

vetted thoroughly before finalizing the model.

6 Epstein, William. Why States Go – and don’t go – Nuclear. The Annals of the American Academy of Political

Science, Vol 430, No. 1, 16-28 (1977) / DOI: 10.1177/000271627743000104

DRAFT

from the academic background of the Technological Imperative, Epstein and others developed this

theory and qualitatively argued that motivations related to the security environment were important to

understanding proliferation. Epstein wrote about security motivation as a driver for nuclear acquisition,

but in line with the technological imperative, he did not differentiate between weapons and energy.

Motive conditions generally outweighed the disincentives, thus states were likely to pursue nuclear

energy (and by default, nuclear weapons) if they faced a security threat or wished to expand their

economy.7

Later, authors evolved to recognize that technological capability was a necessary but insufficient

condition for proliferation to occur. In his groundbreaking work in 1984, Meyer introduced the idea that

having nuclear technology did not necessarily lead to the development of nuclear weapons and that

motivation or the intention of the country’s leaders was also necessary.8 Meyer’s work is also

groundbreaking because he is one of the first authors to move from qualitative research methods like

case-studies into quantitative empirical studies of motivational theories.

Authors continued to ponder the question of proliferation and with the end of the Cold War a significant

paradigm shift occurred in the nuclear weapons world. As the U.S. and Soviet Union stood down their

arms race with new arms control agreements, other changes in the nonproliferation environment

occurred that warranted attention by theorists and undermined existing conceptions of proliferation.

The coalition liberation of Kuwait in 1992 and short incursion into Iraq unveiled just how far Saddam

Hussein had gone towards getting a nuclear weapon. Alternatively, South Africa came clean and

disarmed itself of nuclear weapons under the supervision of the International Atomic Energy Agency

(IAEA). States in the Former Soviet Union returned possession of Soviet-era nuclear weapons to the

Russian Federation and the world focused on dealing with the challenges of secretive nuclear weapons

program development. In 2006, North Korea changed the paradigm even more by testing its own

nuclear weapon following its 2003 withdrawal from the NPT. All of these activities challenged traditional

understandings of the mechanics of proliferation or the capabilities required to proliferate.

Authors like Sonali Singh & Chris Way,9 and Dong-Joon Jo & Eric Gartzke10 attempted to work through

these challenges by developing large datasets of social indicators and performing quantitative tests

using statistical regression. Additionally, they made these datasets available to the public for additional

research. Their datasets broadly cover many of the factors political scientists believe are connected with

proliferation.

7 Epstein, 24

8 Meyer, Stephen. The Dynamics of Nuclear Proliferation. University of Chicago Press: Chicago, 1984. Pg. 90

9 Singh, Sonali and Christopher Way, The Correlates of Nuclear Proliferation: A Quantitative Test, Journal of Conflict

Resolution, 2004; Vol 48, No 6. 859-885. DOI:10.1177/0022002704269655

10 Dong-Joon Jo and Erik Gartzke, Determinants of Nuclear Weapons Proliferation, Journal of Conflict Resolution,

2007; Vol 51, No 1. 167-194

DRAFT

At the same time, researchers like Scott Sagan and Maria Rublee continued advancing qualitative

research about social factors related to proliferation. To help summarize the vast amount of research to

date on social factors of proliferation, we have included a table of information in Annex I.

In a special edition of the Journal of Conflict Resolution devoted to proliferation analysis in the spring of

2009, Montgomery and Sagan identified challenges with predicting proliferation while acknowledging

the quantitative work of Singh & Way and Jo & Gartzke. They noted that there are still challenges with

scarce open-source data about proliferation activities and with coding complex factors such as prestige,

bureaucratic influence and the nonproliferation regime. Additionally, they note that some findings

provide only insights that are already known. Furthermore, the current quantitative literature often

ignores or glosses over data crucial for policy making and wider debates.11

In this environment, the PNNL research team first explored the utility of social modeling for proliferation

assessment, and then built a preliminary predictive model that utilizes a Bayesian Network (BN)

approach, datasets from existing research and structured expert judgment.

BN Models

Bayesian network (BN) models can be used to provide a probabilistic assessment of competing

hypotheses in light of evidence that is relevant to the hypothesis but not necessarily conclusive. BNs are

graphical representations of relationships among variables and provide generalized, quantitative

modeling and analytic capability. Nodes of BNs represent uncertain variables and connections

between nodes indicate probabilistic dependence between variables. Historically, BNs were built using

expert judgment. Learning algorithms are now available that allow models to learn both the BN

structure and probabilities from datasets, thus it is now possible to build BNs using a combination of

expert judgment and available data. The graphical nature of BNs makes their understanding intuitive

and they lend themselves to interactive exploration of alternative hypotheses and the strength of causal

relationships. BNs have been used in medical diagnosis, decision support for the military, intelligence

analysis, features of Microsoft software, and many other applications.

BNs can be used to test hypotheses using causal reasoning, as opposed to identifying correlations

between variables that may or not be causal. Consider the following example case using the BN shown

in Figure 1 intended to help diagnose a problem with a car that will not start.12

11

Montgomery and Sagan, 3

12 Russell, Stuart and Peter Norvig, 2003, Artificial Intelligence: A Modern Approach, Second Edition, Prenice Hall,

New Jersey.

DRAFT

Figure 1 - Simple BN to illustrate diagnostic reasoning13

This figure shows two types of connections (convergent and divergent) and the associated BN reasoning.

Consider the three nodes in the upper left. This exemplifies a divergent connection. If the condition of

the battery is unknown, then learning whether the radio works provides information about the ignition.

Once the battery state is known however, the Radio and Ignition become independent. “Ignition,”

“Starts,” and “Gas” form a convergent connection. Here the situation is just the opposite. Knowing the

car won’t start makes ignition and gas dependent. Knowing the state of one variable changes the

probability of the other being the problem, by “explaining away” the cause.

To see how this type of reasoning can be applied to reasoning about proliferation, consider the model

shown in Figure 2, which represent a simple BN model of proliferation analysis. This BN is a simple

model of conceptual information as it relates to identifying a proliferating state.

13

Ibid.

DRAFT

Figure 2 - Simple Bayesian Net for Reasoning about Proliferation.

The three nodes represent three areas of proliferation analysis that are carried out independently of

each other. The “proliferation detection at a declared facility” node is primarily addressed by material

protection control and accounting (MPCA) and State Systems of Accounting and Control (SSAC). The act

of detecting proliferation at a declared facility depends on determining whether weapons grade

materials are either being diverted from their peaceful purposes or misused. The focus is on

understanding and countering possible diversionary or misuse tactics that could take place at known

facilities. The “proliferation at a hidden facility” node concerns the ability to detect an undeclared

facility intended for the production of weapons grade material such as highly enriched uranium (HEU).

The relevant approach for detecting a facility falls to intelligence agencies and utilizes technologies like

signal detection and remote sensing. The “state intends to proliferate” node is concerned with

understanding the motivations, technical and economic capability of nation states with regards to

weapons production. The technical skills and expertise of analysts in these three areas of endeavor are

very divergent: Political science, nuclear science (physics), geosciences (geography) and economics.

These nodes represent three pieces of the proliferation puzzle combining analysts with different

technical backgrounds and a common interest in the detection and minimization of proliferation, but

who normally do not work with each other.

The BN representing the results of these three activities shows their interconnectedness. Specifically, if

the country’s intention to proliferate is unknown, knowledge of proliferation activities at a declared

facility suggests an increased likelihood of proliferation at a hidden facility and vice versa. Furthermore,

knowledge that the intention of the country is to engage in proliferation activities increases the

likelihood of proliferation at a known facility, which provides insight to analysts working in both of those

areas.

DRAFT

A simple example helps understand this structure and reasoning. Political scientists have long believed

that Iran’s intention is to proliferate.14 Questionable behavior and inconsistencies at their declared

facilities like providing misleading information and appearing to conceal information led analysts to

believe that there could be proliferation behavior underway at an unknown facility.15 In 2009, open

source reports indicated that Iran had a secret facility dedicated to enrichment and located on a military

base. This example shows why this type of reasoning can help identify unknown information (in this

case, a secret facility) by using known information.

The authors have focused on the top node. In particular we have been developing BN models to make

assessments of a state’s intention to proliferate. A preliminary version of a country level proliferation

model is shown in Figure 3.16

Figure 3 – Simple Country Proliferation Model17

14

Perry, Major Richard M. Rogue or Rational State?: A Nuclear Armed Iran and US Counter Proliferation Strategy.

Research Department: Air Command and Staff College, 1997

15 Cordesman, Anthony H. and Adam C. Seitz, Iranian Weapons of Mass Destruction: The Birth of a Regional

Nuclear Arms Race? Center for Strategic and International Studies, Praeger Security International: Santa Barbara,

2009, Pp 3, 188-205

16 Coles GA, SE Thompson, ZN Gastelum, and AJ Brothers. 2009. "Utility of Social Modeling for Proliferation

Assessment - Preliminary Findings ." 2009 International Nuclear Materials Management conference, Tuscon, AZ.

PNNL-SA-64468.

17 Coles, et. al

DRAFT

This simple, illustrative model represents our early conceptual thinking about state level nuclear

proliferation. Despite its simplicity, the model captures many elements of the basic causal structure that

subsequent exploration of the literature uncovered. For a state to proliferate there must be two

elements present: the ability to do so (“nuclear capability” (NC)) and the desire or motivation

(“motivation”). The motivation is in part due to a state’s external political climate. Are its neighbors

capable of launching a nuclear attack? Are they peaceful? How interdependent is the state economically

and diplomatically? Since this model was developed, the authors have identified additional information

that affect “motivation,” such as the aspirations of its leaders to be a nuclear power or the need to

create a national focus to divert attention from a domestic political problem. The other major element,

“Nuclear capability” consists of the inputs “technical capability” and “material resources.” “Material

resources” include economic capacity in addition to the needed materials for a nuclear weapon.

The causal structure of the model is one aspect of BN development. The other is the quantification of

the probabilities. This model was quantified using expert judgment based on factors identified by the

authors in a literature review. The project team performed a preliminary assessment estimating

parameters for 192 countries and made judgments about stability and technical resources in those

countries and their regions to populate probability tables in the model. Our intention was to augment

the model by gathering better, more robust data about the social factors to support or modify these

judgments. Conceptually, this simple model does surprisingly well and provides reasonably accurate

results. Rather than developing this simple model, the team elected to pursue a different route and built

a model incorporating data and structure from existing statistical studies, which are supplemented by

expert judgment as needed. Figure 5 shows the result of exercising the simple, model on two countries,

a low risk country and a high risk country and is provided to help explain the basic causal structures of a

BN proliferation model.

DRAFT

Figure 5 - Example Assessments for simple country model.

In this example, the dependent variable, “Will state dev. Nuc weap. prgrm?” is likelihood, based on the

inputs (all of the other nodes). We can see that the state with a high likelihood to proliferate has nuclear

capability, is isolated diplomatically and economically from the international community, has a

perceived threat from its neighbors and lives in an unstable region. Alternatively, the state with a low

likelihood of proliferating may still have a nuclear capability and nuclear capable neighbors, but has high

regional stability and is diplomatically and economically integrated into the international community.

The model needs calibration to improve the magnitudes, but in a general sense, the numbers are in the

right direction. More detailed exercising of the model reveals behavior that is consistent with

expectations. A comprehensive model incorporating our results from this simple exercise with data and

structure from other authors is under construction and discussed in the next section.

BNs have another useful analytic capability. Obtaining accurate information to input into the model

about a country can be costly, and in some cases impossible. One might ask which information is most

important for predicting proliferation. The BN can assist with identifying highly diagnostic information

that will reduce the uncertainty of the model’s output by calculating the diagnosticity of each variable

DRAFT

relative to the other variables. It is one way to rank the variables in terms of predictive importance and

potentially to prioritize information gathering and input into the model.

PNNL’s Social Factors Model

PNNL is developing a social factors model for assessing proliferation likelihood based on existing

datasets and supplemented by expert judgment. The model reflects proliferation concepts underlying

the Jo & Gartzke and Singh & Way datasets and the table of factors identified by the research team

during the course of the literature review (See Appendix I).

In the model, the final output is a state’s likelihood to proliferate given the known information and

historical precedent. The model utilizes proliferation datasets, including those developed by Jo and

Gartzke and Singh and Way when the data are available. Due to the limitation of finding and coding data

and the limited sample size of proliferating countries, expert judgment is used to supplement the

datasets when necessary. Two nodes, “capability” and “motivation” feed directly into the final node (a

state’s likelihood to proliferate). “Capability” addresses technical capability and resources. “Motivation”

covers incentives and constraints like domestic conditions, security environment and NPT membership.

No dataset or statistical work have been completed to explain the relationship between capability and

motivation, but based on initial work by the authors, a simple risk model was utilized to determine the

relationship between “capability,” “motivation” and “proliferation likelihood.” This preliminary social

factors model is still under development.

Next Steps

The social factors model is currently under further development and calibration,18 which will be followed

by validation. Validation will consist of quantifying19 the model for selected country proliferation

scenarios and comparing the results to what is known or what is believed to be true. The scenarios will

be both contemporary and historical case-studies of states to confirm that the model produces good

predictions. After the model is properly calibrated, the project team hopes to identify users who might

be interested in developing a tool based on the model.

Following further development, this model could be used as a predictive tool to identify proliferation

risk, which might factor into a number of different kinds of policy decisions. The likelihood that a state is

proliferating may increase or decrease based on external and domestic factors which change over time

and might be monitored by using such a tool.

As a policy support tool, with additional development and refinement it could be used in roles as diverse

as developing effective sanctions, supporting the right development efforts, or prioritizing states for

18

Whitney PD, and SJ Walsh. 2010. "Calibrating Bayesian Network Representations of Social-Behavioral Models."

In Social Computing, Behavior Modeling and Prediction. LNCS 6007, p338-345

19 In this instance, “quantifying” refers to the act of putting in information about a scenario. Previous discussion

about “quantifying the model” has referred to the act of having the BN learn the relationships and probabilities

associated with the data.

DRAFT

nuclear assistance. It could be used for sanctions and development by identifying the right elements of a

state’s social conditions to support or deny. In this case, the model would enable sensitivity analysis to

identify key factors that influence an outcome. This key information could help policy makers make

more informed decisions regarding where and when to apply sanctions or aid. For example, if a state

exhibits characteristics representing a state with high technical capability and a perceived high risk from

a nuclear neighbor, it may prove valuable to offer that state support in the form of nuclear defense

guarantees to offset their perceived security risks. There are many trade-offs to that decision requiring

additional analysis. A tool like the one proposed would allow policy-makers to test and determine

potential outcomes associated with certain policy decisions and make more informed judgments. As a

prioritization tool for nuclear assistance, the model could help identify relatively high-risk states to focus

resources on and maximize return on investment for nonproliferation policies. For example, if there are

two states with research reactors containing highly enriched uranium and the U.S. and its allies only

have sufficient resources to convert one of those reactors, how does it make that decision and support it

to Congress and our partners? A quantitative model identifying proliferation risk can support that

decision in a transparent and consistent way. If one state has a higher proliferation risk based on social

factors, then that state could be prioritized over the lower risk state.

Conclusion

At this point, the PNNL BN model and its validation remain a work in progress. Nonetheless, the

experience of the modeling team is that BN models do provide a transparent intuitive assessment tool

that is easy to understand, adapt and to use to perform “what-if” analyses. The hope is that developing

such a model can provide insights not yet obvious from other types of analysis and will be a useful

addition to the analytic toolbox.

With a BN model, the project team will be able to analyze questions not yet addressed in the existing

quantitative research such as the effect of possible correlation between factors. This model could

provide the basis for a tool useful to policy makers to help support decisions and implement policies to

prevent additional proliferation of nuclear weapons.

There remains a great deal of work to be done in this field, and like Montgomery and Sagan note in their

2009 article, “As with much good research, they *researchers in this field] provoke as many questions as

they answer.”20

20

Montgomery, Sagan, 21

DRAFT

Annex I – Table of Proliferation Theories

Source Technical

Capability Factors

International

Security Factors

Domestic Politics

Factors

National Identity

& Psychology

Factors

The Correlates of

Nuclear Proliferation

– A Quantitative

Test21

(Singh and

Way)

Latent capability

Economic capacity

Security threat

Security alliance

Regime type

Political change

Economic openness

Economic change

Symbolism

Why Do States Build

Nuclear Weapons?22

(Sagan)

Latent capability Security Threat Activism (for and against)

Symbolism

Institutional isomorphism

Determinants of

Nuclear Weapons

Proliferation23

(Jo

and Gartzke)

Latent capability

Technology diffusion

Economic capacity

Security threat

Security alliance

Diplomatic isolation

Regime type

Domestic unrest

Regional or global status

International norms

Taking Stock of the

Nuclear

Nonproliferation

Regime: Using Social

Physiology to

Understand Regime

Effectiveness24

(Rublee)

Activism (for and against)

Consistency

Linking with established values

Psychology of

Nuclear

Proliferation25

(Hymans)

National identity consensus

21 Singh, Sonali, The Correlates of Nuclear Proliferation: A Quantitative Test, 2004:24:859.

22 Sagan, Scott, Why Do States Build Nuclear Weapons?, International Security, Vol. 21, No. 3 Winter 1996/97, pp. 54-86

23 Dong-Joon Jo, Determinants of Nuclear Weapons Proliferation, Journal of Conflict Resolution, 2007;51;167.

24 Rublee, Maria Rost, Taking Stock of the Nuclear Nonproliferation Regime: Using Social Physiology to Understand Regime Effectiveness,

International Studies Association, 2008: 10: 420-450

25 Hymans, Jacque E. C. The Psychology of Nuclear Proliferation: Identity, Emotions, and Foreign Policy. Cambridge: Cambridge University Press,

2006.

DRAFT

Source Technical

Capability Factors

International

Security Factors

Domestic Politics

Factors

National Identity

& Psychology

Factors

Why States Go – And

Don’t Go - Nuclear26

(Epstein)

Security threat

Security alliance

Weapons superiority

Nuclear hedging

Legal barriers

Symbolism

International norms

Regional status

International status

Economic status

The Politics and

Technology of

Nuclear

Proliferation27

(Mozley)

Security threat

Diplomatic isolation

Regime type Regional status

Nuclear Tipping

Point – Ch 128

Technological difficulty

Security threat

Bargaining tool

Activism (for and against)

Legitimacy of government

Symbolism

International norms

Tipping Point – Ch

229

Technology diffusion

Security alliance

U.S. foreign and security policy

Domestic unrest International norms

Tipping Point – Ch

330

Latent Capability

Technology diffusion

Materials diffusion

Security threat

Security alliance

Domestic unrest

Hedging

Nuclear-phobia

International norms

Tipping Point – Ch

431

Security threat

Trade sanctions

Activism (for and against)

Leadership

International norms

Symbolism

26 Epstein, William. “Why States Go – And Don’t Go – Nuclear.” The Annals of the American Academy of Political and Social Sciences: Nuclear

Proliferation: Prospects, Problems, and Proposals. Vol 430, March 1977, pg 16-28.

27 Mozley, Robert F. The Politics and Technology of Nuclear Proliferation. Seattle: University of Washington Press, 1998.

28 Reiss, Mitchell B. “The Nuclear Tipping Point: Prospects for a World of Many Nuclear Weapons Sates.” The Nuclear Tipping Point: Why States

Reconsider Their Nuclear Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution,

2004.

29 Campell, Kurt M. “Reconsidering a Nuclear Future: Why Countries Might Cross Over the Other Side.” The Nuclear Tipping Point: Why States

Reconsider Their Nuclear Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution,

2004.

30 Einhorn, Robert J. “Will the Abstainers Reconsider? Focusing on Individual Cases.” The Nuclear Tipping Point: Why States Reconsider Their

Nuclear Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution, 2004.

31 Einhorn, Robert J. “Egypt: Frustrated but Still on a Non-Nuclear Course.” The Nuclear Tipping Point: Why States Reconsider Their Nuclear

Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution, 2004.

DRAFT

Source Technical

Capability Factors

International

Security Factors

Domestic Politics

Factors

National Identity

& Psychology

Factors

propensity for nuclear weapons

Tipping Point: Ch 532

Security threat

Diplomatic isolation

International norms

Regional status

Tipping Point – Ch

633

Economic capacity Security threat

Security guarantee

Political change

Economic change

Government legitimacy

Tipping Point – Ch

734

Security threat

Security alliance

Relationship with U.S.

International norms

Nationalism

Tipping Point -

Conclusions35

Technology diffusion

Support from advanced states

Security threat

Security alliance

Relationship with the U.S.

Regime type Symbolism

Meyer Economic Capacity

Diplomatic isolation

Int’l legal commitments

Preemptive diplomatic intervention by major power

Domestic unrest

Recent, major military defeat

Risk of unauthorized seizure

International status

Regional status

International norms

Peaceful reputation

Kwon Technological Capability

Economic Capacity

Safeguards

Physical Protection

Security threats

Security guarantee

Diplomatic isolation

Transparency

Domestic unrest

Activism (for and against)

Leadership propensity for nuclear weapons

International norms

Duration of Participation in NPT

Linking with established values

32 Laipson, Ellen. “Syria: Can the Myth be Maintained without Nukes?” The Nuclear Tipping Point: Why States Reconsider Their Nuclear Choices.

Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution, 2004.

33 Lippman, Thomas W. “Saudi Arabia: The Calculations of Uncertainty.” The Nuclear Tipping Point: Why States Reconsider Their Nuclear

Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution, 2004.

34 Fuerth, Leon. “Turkey: Nuclear Choices amongst Dangerous Neighbors.” The Nuclear Tipping Point: Why States Reconsider Their Nuclear

Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution, 2004.

35 Campbell, Kurt M., and Einhorn, Robert J. “Avoiding the Tipping Point: Concluding Observations.” The Nuclear Tipping Point: Why States

Reconsider Their Nuclear Choices. Kurt M. Campbell, Robert J. Einhorn, and Mitchell B. Reiss (ed). Washington D.C. : The Brookings Institution,

2004.

DRAFT

Source Technical

Capability Factors

International

Security Factors

Domestic Politics

Factors

National Identity

& Psychology

Factors

Export control Public influence (internalization)

Regional status

International status

Diplomatic Activities (to prove compliance with int’l norms)

Sweeney Economic Capacity

Latent capability

Economic capacity

Security threat

Enduring Rivalry

Regional Proliferation

Diplomatic Isolation

Bargaining tool

Security alliance

Preemptive diplomatic intervention by major power

Int’l Legal commitments

Institutional inertia +

Political change

Risk of unauthorized seizure

Economic change

Economic openness

Activism (for and against)

International status

Nationalism

Peaceful Reputation

International norms

Yim (Note: This is for

90% certainty and

combines results

from explore, pursue

and acquire)

Safeguards

Latent capability

Economic capacity

Population size

Enduring rivalry

Frequency of disputes

Security alliance

Economic change

Regime type

Economic openness

Political change

International norms