Topics in Time Series Econometrics

点击翻译按钮获取中文摘要

Exchange rates are central to the functioning of market economies. The value of the U.S. dollar in terms of currencies such as the Euro and the yen is important not only for trade balances and capital flows, but also for inflation, monetary policy, production and employment. Better understanding of exchange rates will thus help us better understand the behavior of variables that are central to economic welfare and economic policy. Exchange rates fluctuate in response to expectations about future economic conditions. This is captured by a "present value model." In such a model, expectations are discounted back to the present, with expectations of economic events in the distant future having less influence on current exchange rates than do expectations of events occurring in the near future. It is difficult, however, to establish strong statistical links between exchange rates and observable fundamentals that are supposed to well reflect prospective economic conditions. Indeed, changes in exchange rates are very difficult to predict. And insofar as such changes can be linked to fundamentals such as the difference between interest rates in the U.S. and abroad, the link seems counterintuitive, and difficult to rationalize. This research develops and applies econometric results that are appropriate for interpreting exchange rates and other variables that fluctuate in response to expectations about future economic events. The research improves our ability to make probabilistic and statistical statements about the behavior of exchange rates and other variables presumed to be determined by a present value model; improves our understanding of certain well known regularities in the relationship between exchange rates and other variables, including the relationship between exchange rates and cross-country interest differentials; and improves our understanding of the economic determinants of exchange rates. This research also develops statistical and econometric tools for forecasting. Prediction is central to scientific analysis and to sound policy design. One aspect of this work is motivated by what seems to be relatively good empirical success of panel data prediction of exchange rates. (In "panel data" prediction, a researcher attempts to simultaneously predict several variables, in this case several exchange rates.) A second aspect is on improved inference about differences in predictive ability across a set of competing models. This aspect of the research develops and applies econometric results applicable to comparison of nested models. These results help other users of probabilistic models identify good ways to predict. Such users include not only economists, but other scientific researchers and policy makers as well.