The foreign exchange rate is one of the most important asset prices in the international financial market. My dissertation studies the determination of exchange rates from the perspective of levered financial institutions and the frictions they are facing. It consists of two chapters that shed light on the importance of levered financial institutions in exchange rate determination. In Chapter 1, I propose an intermediary-based explanation of the risk premium of currency carry trade in a model with a cross-section of small open economies. In the model, bankers in each country lever up and hold interest-free cash as liquidity buffers against funding shocks. Countries set different nominal interest rates, while low interest rates encourage bankers to take high leverage. Consequently, bankers’ wealth drops sharply with a negative shock. This reduces foreign asset demand and leads to a domestic appreciation, which in turn makes low-interest-rate currencies good hedges. The model implies covered interest rate parity deviations when safe assets differ in liquidity. The empirical evidence is consistent with the main model implications: (i) Low-interest-rate countries have high bank leverage and low currency returns; (ii) the carry trade return is procyclical with a positive exposure to the bank stock return; and (iii) the carry trade CAPM beta increases with the stock market volatility. In Chapter 2 (joint with Yang Liu), we study how time-varying volatility drives exchange rates through financial intermediaries’ risk management. We propose a model where currency market participants are levered intermediaries subject to value-at-risk constraints. Higher volatility translates into tighter financial constraints. Therefore, intermediaries require higher returns to hold foreign assets, and the foreign currency is expected to appreciate. Estimated by the simulated method of moments, our model quantitatively resolves the Backus-Smith puzzle, the forward premium puzzle, the exchange rate volatility puzzle, and generate deviations from covered interest rate parity. Our empirical tests verify model implications that volatility and financial constraint tightness predict exchange rates.