Some Aspects of a Decision-Making Model in Nuclear Deterrence Theory

Abstract

This paper partly develops and analyzes the expected- value model of nuclear deterrence formulated by Daniel Ellsberg, formerly of the RAND Corporation, in his paper 'The crude analysis of strategic choices.'

In the model, the Soviets and Americans each have the choice of Strike (all-out nuclear attack) or Wait (representative or best alternative to Strike) toward the other. Some of the variables of interest are

q = Soviet subjective probability of a US Strike,

q = the Soviet subjective probability of a US Strike at which the Soviets are indifferent between Strike and wait,

V₁₁ = Soviet no all-out war payoff, i.e. payoff for a Wait strategy by each side,

V₂₁ = Soviet first strike payoff,

V₁₂ = Soviet second strike payoff

(where q, V₁₁, V₂₁, V₁₂ are US estimates),

M = Utility (Wait) - Utility (Strike) = (V₁₁ - V₂₁)- q(V₁₁ - V₂₁).

Deterrence holds if and only if M > 0.

The prime indices of deterrence stability are M and q. Cases are presented showing the conditions under which M and q increase simultaneously, or one in creases while the other remains constant, since such increases imply that deterrence is enhanced.

In general such cases require so much information that (1) one is unlikely ever to know if the cases apply, even if they do- -hence predicting whether deterrence is enhanced or not is very difficult; and (2) they could not be employed by a policy-maker seeking to mani pulate variables to enhance deterrence stability.

Also, interesting numerical real-life ranges of changes in values of V₂₁, V₁₂ and q are listed. These ranges of changes lead to both increases and decreases in M and q, unfortunately, so the model once more is seen to be poor for predicting whether deterrence is enhanced or not.

Typical real-life force changes are presented next. In general, one cannot get enough information to be able to predict the consequences for M and q. MIRV and the ABM are dealt with briefly.

The work in this article substantially advances knowledge about this deterrence model, and extends the criticism made by a number of writers who stress that deterrence theory requires rational behavior. The extreme limitations on one's ability to make predic tions even while employing a rigorous model are in vestigated. The fundamental conclusions is that the model does not predict deterrence outcomes for vital, realistic force changes.