Garber on Speculative Attacks 

"The seminal articles in the speculative attack literature were Salant and Henderson's (1978) study of attacks on buffer stocks held to peg the real price of gold and Krugman's (1979) study of attacks on fixed exchange rate regimes...The departure in Salant and Henderson from the previous exogenous regime switching models lay in their realization that the timing of the regime switch was endogenous and that the buffer stock would not drop continuously to zero at the time of the switch but would be attacked and forced discontinuously to zero. At the time of the switch in regime, there would be a jump in the continuous rate of capital gains on holding gold---the rate of change of the real gold price would jump from zero to a positive value. This would now make it worthwhile for private speculators to hold gold bullion, and they would move to acquire gold left in the buffer stocks...Thus, the attack must come when the buffer stock still has gold holdings---indeed, the attack should occur at exactly the moment that the transfer of the remaining stock into private hands would satisfy speculative demand without a price jump. 

These principles---no anticipated asset price discontinuities, endogenous timing of attack on a buffer stock, a discontinuity between pre-attack and post attack rates of capital gain, and the attack's occurring when a finite buffer stock was still in the hands of the authorities---were the concepts that Krugman applied to attacks on fixed exchange rate regimes." 

P. Garber and L. Svensson 
 "The Operation and Collapse of Fixed Exchange Rate Regimes," 
Handbook of International Economics, vol. 3,   p.1893