Implications of within-period timing in models of speculative attack

Speculative attacks are often modeled as decreases in money demand before currency crises. I discuss how, in models with microfoundations, within-period timing affects whether attacks arise in equilibrium. "Cash-when-I'm-done" timing always generates attacks, but is controversial because it assumes that end-of-period money balances buy current consumption. Cash-in-advance timing, theoretically more appealing, generates attacks only under restrictive assumptions. These issues arise when money is introduced via liquidity constraints, the utility function, or a transactions technology. Modeling attacks via reductions in demand for domestic bonds, instead of reductions in money demand, helps avoid these issues, and may be more realistic.