This dissertation studies alternative models for the determination of the real exchange rate level, its volatility and its variations. The first chapter, "Purchasing Power Parity and Breaking Trend Functions in the Real Exchange Rate," introduces a long-run model for the real exchange rate that satisfies purchasing power parity and allows for the Balassa-Samuelson effect. The specification is such that the real exchange rate has a long-run deterministic trend with the possibility of breaks. This specification is estimated by applying an improved framework to test for unit roots and structural breaks Observed data are compatible with the model because, in 19 out of 20 countries, real exchange rates are stationary around a deterministic trend and structural breaks are identified in 12 countries.

The literature on exchange rates has recently explored asset-pricing models as an alternative approach to understanding the determinants of exchange rate returns. The second chapter, "Habits, Catching Up with the Joneses and International Risk Sharing," calibrates a consumption-based asset pricing model such that it is consistent with stylized facts on international risk sharing, exchange-rate volatility, equity-return volatility, the average equity premium and the average risk-free rate. The main contribution of this exercise is that, by including both internal and external habits in the utility function, it is possible to reconcile international risk-sharing measurements and real exchange-rate volatility. The calibrated model implies highly correlated stochastic discount factors and thus high levels of international risk sharing in line with evidence with asset market data from the G7 countries.

The third chapter, "A Consumption-Based Approach to Exchange Rate Predictability," derives a predictability equation for the real exchange rate using the model presented in the previous chapter. Consumption growth can predict real exchange rate returns under the presence of habits in the utility function. This implication is empirically assessed by computing predictability tests with bilateral real exchange rate and consumption data from 17 OECD countries. The results show that, in 15 countries, forecasting real exchange-rate returns with this consumption-based specification produces smaller forecasting errors than with a random-walk model.