Defense policies against currency attacks: on the possibility of predictions in a global game with multiple equilibria

This paper studies defense policies in a global-game model of speculative currency attacks. Although the signaling role of policy interventions sustains multiple equilibria, a number of novel predictions emerge which are robust across all equilibria. (i) The central bank intervenes by raising domestic interest rates, or otherwise raising the cost of speculation, only when the value it assigns to defending the pegits type"is intermediate. (ii) Devaluation occurs only for low types. (iii) The set of types who intervene shrinks with the precision of market information. (iv) A unique equilibrium policy survives in the limit as the noise in market information vanishes, whereas the devaluation outcome remains indeterminate. (v) The payoff of the central bank is monotonic in its type. (vi) The option to intervene can be harmful only for sufficiently strong types; and when this happens, weak types are necessarily better off. While these predictions seem reasonable, none of them would have been possible in the common-knowledge version of the model. Combined, these results illustrate the broader methodological point of the paper: global games can retain significant selection power and deliver useful predictions even when the endogeneity of information sustains multiple equilibria.