摘要 : Although probably reluctant to admit it, many empirical macroeconomists have had the uneasy experience of estimating models, publishing results of these models, and then upon later re-estimation with updated data, finding that the... 展开 Although probably reluctant to admit it, many empirical macroeconomists have had the uneasy experience of estimating models, publishing results of these models, and then upon later re-estimation with updated data, finding that the initial results nolonger hold. In many instances, such problems could have been avoided if the researcher had investigated the robustness of the initial results with different data vintages. Until now, the implicit cost-benefit analysis done by most researchers has suggested that cost of constructing even a few such vintages is overwhelming. In an ambitious and important research agenda, Tom Stark and Dean Croushore have constructed, and begun to evaluate, a real-time dataset. This dataset contains vintages from November 1965 to the present, with each vintage beginning in 1947:1. Each vintage includes key quarterly data from the National Income and Product Accounts and monthly data on important monetary and reserves aggregates as well as the unemployment rate, the CPI.and two interest rates. Each vintage is constructed in the middle of the quarter, and includes only data known at that time. 收起