Abstract: This dissertation consists three studies about floating exchange rates using both theoretical and empirical approaches: (1) forecasting exchange rates using a time-series model, (2) examining interaction between net foreign asset positions and exchange rates under a general equilibrium framework, and (3) a comprehensive survey on theoretical and empirical approaches to exchange rate puzzles. In the first essay, I construct a new forecasting model for exchange rates: the Markov-Switching Error-Correction (MSEC) model. In this model, an unobserved Markov process captures the short-run dynamics of exchange rates, while an error-correction term describes the long-run equilibrium in foreign exchange markets, which is driven by macro-fundamentals. Over the period from the first quarter of 1996 to the last quarter of 2005, my model provides better out-of-sample forecast performance than the random walk model at one- to twelve-quarter horizons. The second essay analyzes the interaction between net foreign assets position and exchange rate in developed countries. An economy's external position is determined by three factors: assets inflows and outflows, assets prices, and exchange rates. Consistent dollar depreciation gives big attention to the valuation effect of exchange rates. A simulation study in a general equilibrium monetary model with international portfolio choice shows that the valuation effect plays a significant role in determining a country's net foreign assets position. The third essay surveys three exchange rate modeling approaches: new open economy macroeconomics, market microstructure, and model with distorted beliefs. Compared study indicates that new open economy macroeconomic models can help to set up the relationship between exchange rates and macro-fundamentals at long horizons while the market microstructure approach and model with distorted beliefs may help to describe the short-run dynamics of exchange rates.