| The difficulties of modeling and forecasting foreign exchange rates have been well known since early 1970's. One of the possible explanations for our inability to provide an accurate model is the structural changes over time, especially in emerging markets. The traditional regression technique that assume constant parameters are incapable to capture the changing dynamics over time. Consequently, most foreign exchange regression models are ineffective.;To better capture the fundamental structural changes in a market, a moving block regression technique is recommended by the author. The moving block regression procedure utilizes sub-sample information, rather than the prevailing whole sample regression that intends to increase regression efficiency with more observations. To find out the loss or gain of forecast efficiency of the proposed moving block technique relative to the whole sample method, a Monte Carlo study is carried out under several different scenarios: data in compliance with the classic regression assumptions, data with heteroscedasticity, data with autocorrelation, missing variables, changing regression coefficients, and nonlinear data.;Simulation results show a small trade off in presence of minor assumption violations. There is a clear dominance of the moving block technique over the traditional whole sample regression in terms of forecast efficiency when violation of assumptions are serious, such as missing variable, changing coefficients, or nonlinear data.;Then the moving block regressions are applied to six exchange rates against the U.S. dollar. The comparison of forecasting residuals, both in-sample and out-of-sample, shows a strong support for the moving block technique indicating the inevitable violations of regression assumptions in foreign exchange markets. |
