To update an old statistic: a political leader in a developing country is almost twice as likely to lose office in the six months following a currency crash as otherwise. This difference, which is highly significant statistically, holds regardless of whether the devaluation takes place in the context of an IMF program. Why are devaluations so costly? Many of the currency crises of the last 10 years have been associated with output loss. Is this, as alleged, because of excessive reliance on raising the interest rate as a policy response? More likely it is because of contractionary effects of devaluation. There are various possible contractionary effects of devaluation, but it is appropriate that the balance sheet effect receives the most emphasis. Pass-through from exchange rate changes to import prices in developing countries is not the problem: this coefficient fell in the 1990s, as a look at some narrowly defined products shows. Rather, balance sheets are the problem. How can countries mitigate the fall in output resulting from the balance sheet effect in crises? In the shorter term, adjusting promptly after inflows cease is better than procrastinating by shifting to short-term dollar debt, which raises the costliness of the devaluation when it finally comes. In the longer term, greater openness to trade reduces vulnerability to both sudden stops and currency crashes.