| My dissertation consists of two projects. In the first paper, I try to enhance our understanding of preliminary data releases of industrial production (IP) and the composite leading indicator (CLI). There are many examples in applied economics which illustrate the need to take a closer look at questions pertaining to the quality of preliminary data releases. For example, many econometric forecasting models are routinely constructed using "currently available" data. In the U.S., data are often downloaded from CITIBASE, and used without giving too much thought to the "timing" of the data. However, it is well known that many CITIBASE series are formed by combining various different "vintages" of economic data (e.g. preliminary data and data which have been revised a number of times). Several time series properties of this data revision process are examined. In particular, I examine moments, autocorrelation functions, and integratedness properties of the revision process, and also construct and estimate univariate and multivariate regression and forecasting models in order to assess the "efficiency" of IP and CLI revisions.; In the second paper, I examine the prevalence of model and data uncertainty in the formation of simple monetary policy rules which mimic real-time policy-making decisions. By model uncertainty, it is meant that the specification and/or parameters of a model are no longer assumed to be fixed and known. While ignoring this type of uncertainty often leads to tractable models which are easily analyzed, it may also paint a picture of the world which is oversimplified. By data uncertainty, it is meant that first released data are often noisy in the sense that incomplete and/or erroneous initial information has been used in their construction. Indeed, it may take many years of revisions before data are considered final. Furthermore, actual policy decisions are made in a real-time setting using preliminary and/or partially revised data. Thus, questions relating not only to which variables should be used, but also to which data releases should be used, make the process of policy-making much more complex than is typically assumed in abstract models of monetary policy. In this paper these issues are considered by examining monetary policy-rules using adaptive and recursive learning based on least squares, and using data which are available in real-time. The approach is to build real-time datasets, simulate a real-time policy-setting environment, and provide a set of prescriptions and diagnoses which are useful not only within the context on monetary policy rules, but also within the context of the application of real-time data to macroeconomics in general. (Abstract shortened by UMI.)... |
