Three essays on the foreign exchange markets

The foreign exchange market is full of risk and uncertainty. Researchers are trying many approaches to explain foreign exchange movements and forecast foreign exchange rates. The most quoted methods are forward premium model and vector error correction model (VECM).;The first essay study the explanation power of the macroeconomic news for the foreign exchange fluctuation. We use Kalman filter and maximum likelihood method to extract several dynamic factors from 27 noisy and sparsely observed macroeconomic news deviations. We further input the news factors as independent variables in our VECM analysis. The fitted results show that the news factors' contribution is limited. The out of sample prediction yields the same conclusion.;The uncovered interest rate parity hypothesis has frequently been rejected. This hypothesis, however, has seldom been tested at the very short end of the forecasting horizon where forward rates are measured in days. The second paper reinvestigate the UIRP puzzle in diversified horizons. Using overnight, two-day and three-day forward rates, we find that the forward premia in these short forward horizons are stationary than the forward premia in longer horizons. This contrasts with recent findings that the forward premia, in longer forward horizons, are fractionally nonstationary. Estimation results indicate that forward premia are essentially unbiased estimates of the future spot returns. Once the interest rate differential is the dominant source of information in the foreign exchange market, the forward premium forecasts the spot returns relatively well.;The last essay we have a empirical study of the VECM prediction power we use the rolling regression method to generate series of parameters and dynamically predict the next period's foreign exchange rates. We compare the forecast errors from the rolling regression VECM and that of the random walk model. We also set up a trading strategy which longs or shorts the foreign currency based on the forecasts. Our trading simulation shows that this informed trading makes positive return in medium horizon, while the simple buy and hold strategy's return is insignificant.