Executive Summary

Despite the Singapore and Hanoi summit meetings between Trump and Kim Jong-un and inter-Korean summits between Moon Jae-in and Kim Jong-un, there has not been any real progress towards ‘final, fully-verified denuclearization’ over two years. If anything, all signs point to the contrary. At a military meeting in May, 2020, Kim Jong-un was reported to have called for stronger “nuclear war deterrence” by the state media.1 In the following month, North Korea blew up an inter-Korean liaison office in the North Korean city of Kaesong. With the prospect of North Korea’s denuclearization diminishing, it is imperative that ‘deterrence,’ our subject of interest, is robustly maintained against the North Korea’s growing nuclear threat. In particular, we have become interested in Perfect Deterrence Theory developed by Zagare & Kilgour as an alternative to classical deterrence theory. Above all, its predictions are more in agreement with empirical findings and free of irrationality and, therefore, the theory has been chosen to be the basis of our study.2 It has basically provided a framework for exploring South Korea’s deterrence against North Korea. Our role was to understand its solutions3 (i.e., Perfect Bayesian equilibria’) found by Zagare & Kilgour, and to draw out valuable implications in the context of North-South Korea. Our study was limited to Direct Deterrence4 which includes the Generalized Mutual Deterrence Game and the Unilateral Deterrence Game. Our study has drawn extensively on their book “Perfect Deterrence.3

Chapter 1 is a brief overview of Perfect Deterrence Theory. Some advantages of this theory over classical deterrence theory are mentioned. One advantage is that Zagare & Kilgour’s imposition of the ‘Perfectness’ condition prevents us from encountering irrational solutions later on (e.g., classical deterrence theory cannot explain the paradox of mutual deterrence). The theory also explores two differentiated players (i.e., challenger and defender), and gives due consideration to the ‘status quo’ which was much neglected in favor of the cost of conflict. It also provides rationale for minimum deterrence.

Chapter 2 reviews basic concepts in Perfect Deterrence Theory that are required for understanding the later chapters. ‘Capability’ and ‘credibility’ are two basic components which play critical roles in the theory. Capability is one’s ability to hurt one’s adversary while credibility is one’s willingness to fight rather than capitulate. Some examples are provided in the context of Korea. The Korean War (1950-1953) is an example of incapable South being invaded by capable North. The Admiral Yi Sun-sin’s famous saying, “Those who seek death shall live. Those who seek life shall die,” before defeating Japanese fleet at the battle of Myeongnyang in 1597 epitomizes the highest level of credibility. By contrast, the sinking of Cheonan, a South Korean navy corvette, and the subsequent shelling on the island of Yeonpyeong in 2010 could be partly attributed to low credibility. Finally, the definitions of type ‘Soft’ and type ‘Hard’ are given and the incomplete information game is explained using these types.

Chapter 3 introduces the Generalized Mutual Deterrence Game. Our primary interest is in the Sure-Thing Deterrence Equilibrium among all solutions found by Zagare & Kilgour.5 Its existence conditions contain North Korea’s utilities for the ‘Status Quo (NSQ),’ ‘Conflict (NDD+)’ and ‘North Korea wins (NDC).’ We have examined each utility variable and suggested ways to strengthen deterrence (i.e., robustly fulfilling the existence conditions). We caution against, for example, blindly increasing the value of NSQ in the absence of genuine progress in North Korea’s denuclearization. The only sure way to bring about deterrence is to decrease NDD+. This necessitates a show of force including, for example, a display of new high-tech F35A Joint Strike Fighters and the establishment of ready ‘Decapitation Unit.’ Also, South Korea’s credibility (as perceived by North Korea), PS, is also a critical factor in deterrence calculations. Some noticeable failures in the South Korean military in recent times, which contribute to the lowering of PS, are noted. We have reviewed Zagare & Kilgour’s other equilibria (i.e., the Attack Equilibria and the Bluff Equilibrium).5 Deterrence always fails in the Attack Equilibria. In the case of the Bluff Equilibrium, with both Koreas lacking credibility, particular attention should be drawn to the inherent danger of an unwanted war/conflict arising out of mutual bluffing and misjudgments of each other. A scenario in which North Korea’s credibility, PN, relating to its use of nuclear weapons could be genuinely high is also mentioned.

Chapter 4 introduces the Unilateral Deterrence Game in which South Korea, a defender, continues to preserve peace and stability while North Korea, a challenger, seeks to defeat South Korea. This is an asymmetric game. Once again, our main focus is on deterrence equilibrium and, in particular, on the Certain Deterrence Equilibrium among all solutions found by Zagare & Kilgour.6 As a way of increasing North Korea’s utility for the status quo, NSQ, South Korea must avoid funding, in effect, Kim Jong-un and his trusted super-elites’ luxurious lifestyles in Pyongyang. Money can be easily funneled to support North Korea’s nuclear weapons and ballistic missile programs, which undermines deterrence. NSQ is not to be over-trusted as it can change overnight. Instead, our efforts should be focused on decreasing NDD+. Strengthening both active and passive defense lowers NDD+ and, in particular, the U.S. THAAD anti-missile defense system must be upgraded and integrated with the Patriot systems in operation without delay. Despite recent disputes between South Korea and Japan, both countries share the same core values – freedom, democracy, respect for human rights and the rule of law. They must work closely to face up to North Korea’s challenges and beyond. As for increasing PS, South Korean military’s new rules of engagement, empowering frontline commanders to order retaliation swiftly, serves as a good example of strengthening deterrence.

Among other solutions (the Attack Equilibrium, the Bluff Equilibrium and the Separating Equilibrium), the Bluff Equilibrium is particularly noted for its inherent advantage for the defender (i.e., South Korea). The incident of August 20, 2015, in which North Korea exchanged fire with South Korea over loudspeaker, is viewed from this perspective.

In the concluding chapter, the importance of maintaining high credibility, PS, is stressed again as many deterrence failures can be traced to low PS. In Perfect Deterrence Theory, being capable is necessary (but not sufficient) for deterrence to hold. South Korea is not capable of competing against North Korea in the nuclear arena. As the country totally depends on U.S. extended (nuclear) deterrence to defend against the growing North Korean nuclear threat, the U.S. needs to provide concrete assurance to its allies in this region.

Table of Contents

-Executive Summary
1. Overview of Perfect Deterrence Theory
2. Basic concepts in Perfect Deterrence Theory
   2.1 Capability and Credibility
   2.2 Type Soft, Type Hard and Incomplete Information
3. Implications of the Generalized Mutual Deterrence Game
   3.1 The Sure-Thing Deterrence Equilibrium
   3.2 The Attack Equilibria and the Bluff Equilibrium
4. Implications of the Unilateral Deterrence Game
   4.1 Deterrence Equilibrium (Certain Deterrence and Steadfast Deterrence)
   4.2 The Attack Equilibrium and the Bluff Equilibrium
5. Conclusion


본 보고서의 내용은 필자들의 견해로 아산정책연구원의 공식 입장과는 다를 수 있습니다.

  • 1. “Kim Jong-un calls for greater ‘nuclear war deterrence,’” The Korea Herald, 24th May 2020. http://www.koreaherald.com/view.php?ud=20200524000204
  • 2. This author is solely responsible for any errors of fact or misinterpretation of Perfect Deterrence Theory.
  • 3. Frank C. Zagare & D. Marc Kilgour, “Perfect Deterrence,” Cambridge University Press, 2000.
  • 4. ‘Direct Deterrence’ essentially describes a (strategic) situation in which two players face off against each other whereas ‘Extended Deterrence’ describes a (strategic) situation in which one player (defender) is trying to protect its protégé against the other player (challenger).
  • 5. See Appendix 4 of 3 for detailed derivations of these equilibria which are grouped into Class 1, Class 2A, Class 2B and Class 3. Class 1 includes the Sure-Thing Deterrence Equilibrium which preserves the status quo. Besides this, it has the Separating Equilibrium and the Hybrid Equilibrium. See also Table 1 in Appendix at the end of this report. Note that Class 2A and Class 2B correspond to Class 2N and Class 2S respectively.
  • 6. See Appendix 5 of 3 for detailed derivations of these equilibria which are grouped into Certain Deterrence & Steadfast Deterrence, Separating Equilibrium, Bluff Equilibrium and Attack Equilibrium. See also Table 2 in Appendix at the end of this report.

About Experts



김종우 박사는 아산정책연구원의 계량분석센터 선임연구위원이다. 런던대학교에서 이학학사와 임패리얼 컬리지에서 상대성이론 연구로 이학박사 학위를 취득하였으며 캠브리지대학교 컴퓨터학과에서 Diploma 학위도 취득하였다. 유럽 랜드연구소의 Choice Modelling과 Valuation팀에서 연구원으로 재직하였으며 삼성 메모리 반도체 연구소에서 책임연구원으로 활동하였다. 또한 영국의 PCMS-Datafit에서 Java 소프트웨어 개발업무를 담당하였다. 주요 연구분야는 이산선택모델, 그리고 교통, 보건, 통신 및 유틸리티 분야의 Stated Preference 모델 개발, 공공 서비스가치 책정, WTP (Willingness-To-Pay) 등이다. 주요 연구물로는 "Security at What Cost? Quantifying Individuals’ Trade-offs between Privacy, Liberty and Security,” RAND Report (2010)와 “Modelling Demand for Long-Distance Travellers in Great Britain: Stated preference surveys to support the modelling of demand for high speed rail”, RAND Report (2011)외 다수가 있다.